📚┃suno-school

[lapis sable]

I didn't think about that, thanks.

Your song is really good, it was worth the effort

[pure panther]

Thank you 🙂

[sharp kettle]

someone taught me how to prompt suno>
so heres a sample lay out.

Genre: 1954 rock and roll, blues
Mood: danceable high energy and rhythmic
Key instruments: Hand drums, Bass Guitar, Acoustic Guitar.
Vocals: Instrumental (No Vocals).( or). studio quality, clean Vocals
Production: Clean mix, crisp diction, studio quality, dry audio,
play in the style of a 1950s rock & roll pioneer, play in the sound of burning love, Crazy Little Thing Called Love.
add exotic dance hand drums rhythm and a boom bap hand drumbeat plays during the interludes.
Tempo is 140 bpm.
Description: A pristine, studio-quality track, Drums and bass maintain a strong, driving danceable high energy rhythm throughout the song.

edit the lay out to fight yours some thing can be changed, some can not, vocals is for ether instruments only or vocals

[pure panther]

Nice, thank you! You don't find that Suno sometimes sings the prompt using that format? I've used something similar and moved away from it due to that.

[sharp kettle]

[Instrumental Intro: 15 seconds long][1954 rock&roll driving high energy][140 bpm tempo][key Instruments Handdrums, Acoustic Guitar, Bass Guitar] put something that in lyrics if you add this.

[sharp kettle]

Genre: 1954 rock and roll, blues
Mood: danceable high energy and rhythmic
Key instruments: Hand drums, Bass Guitar, Acoustic Guitar.
Vocals: Instrumental (No Vocals).( or). studio quality, clean Vocals
Production: Clean mix, crisp diction, studio quality, dry audio,
play in the style of a 1950s rock & roll pioneer, play in the sound of burning love, Crazy Little Thing Called Love.
add exotic dance hand drums rhythm and a boom bap hand drumbeat plays during the interludes.
Tempo is 140 bpm.
Description: A pristine, studio-quality track, Drums and bass maintain a strong, driving danceable high energy rhythm throughout
the song.
Song lyrics Theme: happy thoughts, birds, drums, singing, hearts...

you will be suprised how ai writes the lyrics

[pure panther]

Oh ha, I write all the lyrics. I've seen my hubby use some funny prompts and get good ones though 😂

[sharp kettle]

me two but sometimes i have trouble with lyrics so i have ai fill in based on the prompt the rewrite once it gives me an idea lol

also no becaus i am pro and i set the follow high and set my lyrics already 😂

[sharp kettle]

how does one prompt the generator to do the mouth sounds to musical struments and composistions.
i kept seeing that ad on Instagram.
and i can mouth rythm for manyy of my songs. I could do that for the sound breaks

lika nah naha nah..

[sharp kettle]

can some one change this prompt keep format so
that is changes my voice and sounds to the intrumenatal beat instead( my PRO creds have refreshed and i want to do the vocal sounds to instrument thing i keep seing on instagram

Genre: 1954, 1964, 1972, 1980. 2025 , rock and roll.
Mood: danceable high energy and rhythmic
Key instruments: Hand drums, Bass Guitar, Acoustic Guitar.
Vocals: Instrumental (No Vocals), studio quality, clean
Production: Clean mix, crisp diction, studio quality, dry audio,
play in the style of a 1950s rock & roll pioneer, play in the sound of burning love Crazy Little Thing Called Love
add exotic dance hand drums rhythm and a boom bap hand drumbeat plays during the interludes.
Tempo is 140 bpm.
Description: A pristine, studio-quality track, Drums and bass maintain a strong, driving danceable high energy rhythm throughout the song.
Use primarily D,G,A, major chords and b minor for an upbeat, arousing and seductive dance melody.

[quiet solar]

I believe most of the Instagrammers are using a different tool which specifically focuses on voice replacement, turning vocal audio into a specific instrument.

Suno can sort-of do something similar with an audio upload, but it's not what it's designed for. Rather than a single instrument output, Suno will typically revert to creating a song (especially if you give it a prompt like yours) but using aspects of the audio you input to influence the output.

If you're happy with that, your prompt is fine. Just beatbox the drums or tap it out clearly with your fingers or whatever improvised instruments you have handy, and Suno should pick that up as an influence for the percussion line of the song. You can then get the individual stems and pull the drumline back out if you're looking to post something back to Insta - although you might need to do some editing in post to clean up the audio quality of the drum stem.

If you just want to generate a specific solo instrument, you want a much simpler prompt focusing specifically on the instrument you're looking to create and asking for an instrumental solo by that instrument and any production effects you're looking for. You might be able to get away with including a genre style, but I wouldn't prompt for anything else.

So in your preferred layout, the style prompt would be something like:

Genre: 1950s Blues
Key instrument: Saxophone
Production: Clean mix with limited production effects
A bold, dreamlike saxophone solo

And then in the lyrics box, you probably want to put something like:

[Instrumental Intro - saxophone solo]
[Instrumental interlude - bright, upbeat saxophone solo]
[Hook - emotive saxophone solo]
etc. etc.

You'll probably still get drums behind it sooner or later whatever you prompt for, so anticipate having to split the instrument stems at the end to get a clean solo.

The musical term for that is the vocals being "forward in the mix", which you've already tried. Prompting for production effects is still a bit of a dark art as far as I can tell, so I'd be tempted to consider it a problem you can only partially solve through prompting.

A non-prompting solution would be to split the vocal/instrument stems, open them in Suno Studio or your DAW/audio editor of choice and change the volume mix of the two stems yourself.

If you search for "duet" or "male and female" in this channel, you'll find a few suggestions for ways to go about setting up duets.

Main thing to know is that it isn't something you can control through prompting alone, and that there are broadly three ways to go about it:

1. Create separate male/female versions and combine them
2. Create a starting version to lock in instrumentals, then use Extend to create vocal stem fixes for any vocal mistakes.
3. Create the song using multiple extensions each time you need to switch singers, optionally with a persona

1 and 2 will require the use of Suno Studio, or an external DAW/Audio editor like Audacity. 3 can be done entirely within Suno on any account.

There are pros and cons to each approach, and which is 'best' may depend a lot on what kind of duet you're doing - how frequently do singers switch over, do they have any harmonies, etc.

1 is great for multiple switches and no harmonies
2/3 are better suited for minimal switches and harmonies

If you want multiple switches and harmonies, I'd personally suggest 2/3 - but know in advance this is basically hard mode, and is going to take quite a lot of time and editing to set up.

[sharp kettle]

https://www.youtube.com/watch?v=9rBSE7TBBdM&t=339s this is straight up what i am trying to do. no just with the tempo meows. but my voice is also in tempo/ rythm for the verses and chorus.... i also saw an add to do this with suno on sunos instagram.... and yes i finally found a video..... maybes its a version 5.0 pro version feature.. but it was cool and i kinda wanted to do it lol

[quiet solar]

Okay, that's basically what I was describing in the 'full song' route. He literally describes the workflow for doing it at about 1m-3m into the video. It's not a v5 or Pro feature - it's all based on using Audio Upload as part of the Create process.

The prompting influences what Suno will try to make overall - the audio you feed in will influence that somehow. Sometimes very dramatically, sometimes just being a particular vocal or musical hook. If the audio sounds like vocals, expect it to be sung. If it sounds like drums, expect it to appear in the drum line, etc.

The Audio Upload tool opens the door to doing some very creative things as well, if you're comfortable using Studio or another editor. Like adding a bridge to a song where your own singing acts as a melody template for the existing AI singer to perform.

[sharp kettle]

what they' re doing never actually happens in suno thats why i am so confused, i cant afford the actual voice to make instruments software because its like something like 2000$ last i checked suno is cheeper.

im music sound trained so i can do meow singing scatting drum tap etc. but no matter the record the beat it never ever hits the sound like you see in the videos. nasical if a drummer heard it the gould drum it, a music guild guitar it. the rythm is supp crips and clear that even the ai
wrote

A novelty song with a male vocalist singing in a baritone range, frequently incorporating cat-like 'meow' sounds, The song features a simple, repetitive melody and a consistent tempo, The instrumentation is sparse, primarily focusing on the vocals, The production is clean with no noticeable effects on the vocals or instruments, The song structure is verse-chorus, with the 'meow' sounds acting as a recurring motif, The key is major, contributing to the lighthearted and playful mood, The tempo is moderate, allowing for clear articulation of the lyrics and 'meow' sounds, There are no complex chord progressions, maintaining a straightforward and accessible feel fow thye meow sound 😂

[quiet solar]

Sure it does. I've had some pretty reliable successes using Audio upload in various ways.

As a random example purely to prove a point, this was made by giving a short audio prompt vocalising the bassline while knuckle and finger tapping for the bass/snare drumbeat: https://suno.com/song/eefab713-01f2-423a-a9dd-cb294d97be05

What I will say is more minimal prompting and handing over more control to Suno often works better with these kinds of inputs, so you might want to experiment with other prompting styles.

I've also had Suno cover fully acoustic recorded songs, which it can be absolutely fantastic at, down to the level of picking up the exact chord and strumming patterns on guitar, and subtle vocal wavers and stresses, and then translating that across to the AI "performers". It just takes a bit of tweaking with the slider settings to find the sweet spot, and a bit of experimentation on prompting to push the specific things you want.

[polar spade]

On top of what @quiet solar is saying, just try this; take any song. Stem out the vocals. Then cover the vocal in a simple clean instrument genre, such as folk guitar. You should get an instrumental of a guitar playing the melody line of what was previously a sung lyrics.

Singing on the audio recording and covering as an instrumental is just the same thing . The more clean the audio, the better though.

Its been a tool for quite some time. Youtubers just need clicks and spend 12 minutes saying something that takes 2 minutes just to hit the algorhythm and get monetised.

That video was 5 months ago as well. We now have Studio that can do it also.

[sharp kettle]

lol ive already sung the songs out what i am trying to do is fill the intro, the interludes and outro with a specific beat.
then have suno ad instrumental to match the exact tone and rythm( or as close as i can to it.). suno struggle with this. and ive seen id advertised like over 40 times... my new audio file is in tone and baat and rythm (used the spectograph to mak it in music bars... so as soon as suno has a spectrograph reader in the tranion my music will read lol.
i also used audactits beat track function as a test audio file. suno couldnt take that test beat and substitute out instruments.

i always end up having to cut each note or segment apart in spectrograph the remix the usable notes.
thats the same as hand playing instruments. im trying to get suno to generate 15 seconds, 30 second and for 1 song a minute beat.
the only cuting i should have to do is cutting out the bad music wierd ness lol but i shour have 5 usable secgments instad of like 40 segments. with maybe 2 usable. im trying audio cut file i just made with mucia the vocals to see if it better matchs next lol

heres an example where it has only 1 interlude but the song is only 1 min and 24 seconds.
but if you listen the vocals already set 3 music factors and the spectrograph is in bars becauase that is how i trained for singing national anthems.
https://suno.com/s/w5oWap7HaAwHMvpo

[radiant oar]

I am not sure I'm totally reading what's going on here my biggest issue I'm a retired musician I've been retired for 9 years I first started using a hammerhead 808 drum machine when I first started learning to compose drums like 25 years ago I learned Mexican shortly thereafter I I forgot at 5:00 I got taken away of course not long after that drums they're very shortly learned afterwards and I can I can compose drums all day but I can't play the style I like to save my life cuz it's too complex it's too fast in my ADHD cannot keep up

The harder I concentrate on it the more I slip off

But I found was soon as soon as best use for people who don't understand music and don't understand music theory cuz you know they speak the emotion they speak the concept they speak the idea into suno and they get a general idea what they want and they kind of have it come out because they don't have this exact idea in their head of what they want to hear they just have this concept of this idea of what they want to sound like it works for them the problem with me is I have an exactly like you you have an exactly I'm starting to get to the point where I can't afford to the premium right now but I'm getting to the point where I kind of need to I need to be able to get the track the Bass track the baseline track just the idea of what I'm getting out look at it see what she knows come up with and then from there separate the stems and I really want to pull the drum track out I want to program the drums myself if I have to do that to start I'm going to use the drum machine or I'm going to use this website I'd rather not use any more AI to be honest with you I'm using Grok try to help me translate what I'm doing into it in a form of sooner is going to understand the only problem is even Grok struggling to get to know to do what it wants it to.

Unfortunately I cannot get enough of an explanation here or the discord is going to allow...

Let me see if I can find a way to make this work..

[sharp kettle]

if suno ai could read spectrograph bars instead of wave forms that almost like litterally sheet music for the tranior to create similar beats to what you want. i was litteraly singing in music bars for that song so it litterally the beat in all instruments.
aadjust the pitch and boom different beat lol

[quiet solar]

Wouldn't it just be simpler to record it as a waveform (ie, just record it) for Suno rather than converting it into a spectral sample to achieve the same effect? Or, vice versa, are you not able to reverse the process and output your spectral sample as a new waveform?

[sharp kettle]

short story i took a music class recently to see if i could learn to read sheaet music. big nop on that)
,but the instructor did show me how to read a spectragraph, its like a sheeat music but its set in bar movements so it only rovides a musician a baseline for playing music if they cant read sheet music becaus its easier to read. the notes can be read almost 1 octave higher or lower. but it shows that artist which beat to hit... it was way easier for me to read the sheet lol. its kinda perfect for current ai because it can create like 4 sounds on one track spectra graph lol

[radiant oar]

You talking to someone that wants to feed it actual sheet music tablature and timing patterns and drums and it just simply doesn't work that way I'm about to start playing my own instruments or actually start feeding other instrumentals in there I mean it okay with the guitars it does kind of okay when I explain to it one of my signature styles that I write with them and then I typically do have the the kick patterns loop at the 16th, 32nd, 64th, or 128th beat

And yes this is called polyrhythm it's influence heavily influenced by like Danny Carrie Neil Peart and Thomas Haake

[polar spade]

Do i understand correctly:

You sang ONLY vocals (solo) of a song. No music, just singing the melody.

You then want to go back and add beats to sections (intro, interlude) either by beat boxing, basically those sections or getting suno to match your vocal attempts and then splice those onto the vocal track

THEN , have Suno create a whole song, adding the production and instrumental track onto the suno-beat/vocal track you have uploaded to Suno or have made previously, so that the instruments, rhythm guitar, etc match the rhythm and beats of the seed song. ?

[radiant oar]

Honest I'm trying to do everything I don't have the money to spend on the equipment I'm using sooner to make up for that it's cheaper to pay suno than pay for all the instruments, and studio time, my engineer

[hollow adder]

Scat

[polar spade]

Yes. Same for me. But its a tool, not a crutch. Theres a place in this world for some studio time. Ultimately, learning as much music editing skills as you can will be beneficial to trying to NOT use a Professional studio. Ai csn only do so much without sacrificing or limiting in other areas.

That said, in many cases, ggood enough is good enough.

[sharp kettle]

the song vocals i provided for that song is in both bars and rhythm. its technicaly the beat for the whole song..
so yeah i would like suno to read those bars and rhythm and fill in the music, it can do its own thing for the melodic parts in the interlude intro and outro haha

[hollow adder]

It’s the artists fault for not naming scales and functions like formulas

🤣

Each person has their own prompt vocabulary it’s tedious

[sharp kettle]

im not that great at scatting... humming. melodic singing, rythm singing and bars sing, and of course tempo 128 - 140 i'm excellent at lol...

[polar spade]

OK. If i can play around and see if I can do that, I'll let you know. Bank says I have to pay my mortgage first, so that experiment will come later.

And dont shoot me if its all done in Studio. Im starting to get spoiled.

[sharp kettle]

its a nations anthem so as long as you use the genre prompt in the lyrics( and the lyrics no edits) you can do your own version 😂

yes I'm New Horizens Nations offical anthem writer lol right now i'm working on a protest march anthem for the lbgtqia+ nation

[polar spade]

I dont feed publicly available music/melody/lyrics back into Suno to avoid poisoning my own songs down the line.

I intend to just make up a quick jingle of 2 verses.

[sharp kettle]

totaly get that espesicaly if you make your own lyrics and music styles

[polar spade]

Yea. Maybe I can just do the "+" version. 🤣 in my corner of the woods that just called "red neck". Or maybe a Teddy Swims vibe.

[sharp kettle]

fun fact + refer to straight allies and the new humans discovered from 2018- to june 2025, and people promopt song about it haha

autism sapain with asd for day winn mwhaha

[quiet solar]

@sharp kettle If I understand right, part of the challenge you're running into is that you're feeding spectral data into Suno via Audio Upload and that's why it's not working?

If you're struggling with a workflow for how to add in a new intro, interlude and outro to an existing song with a specific beat created from an audio prompt, I can give that to you. But you'll have to give Suno audio in a form it can use, or it's not worth me typing it out. 😉

Let me know if that workflow would be useful. Otherwise, I think I've exhausted the advice I can give you. Best of luck!

[shy badge]

Hey all.

This is a Google Gemini Gem. Editable as well so you have access to the gem prompt to change/use in other AI platforms.

Use is simple.

1. Input or paste lyrics you want.
2. Provide styles/genre etc.

Suno Gem does the rest. However, you have access to the source.

Creates great results.

https://gemini.google.com/gem/1fasDFlkGZKfr3J3eIxSPBGTAC2jEOfyk?usp=sharing

[sharp kettle]

lol i keep breaking sunos ai voice cover, and parts of its instrument generation the song i have the most versions of is very complex because it mix musix from 3 generations. 1954- 1972 Rock& roll( which suno thinks is rock), dance rock from 1980s( again sun suno thinks its rock) and exotic Dance( which suno sadely thinks is r&b, or or salsa)(its actuall a fusion music genre)which was create in the 1980s from 3 different music genres,
so i will just keep breaking th model til it learns. but yeah if a later version does com as pro if i got a great generation i would like to regen that version again untill it perfect lol

[candid cliff]

Hi everyone, I was logged out of the app and when I went back in all my songs disappeared. Can someone help me please?

[polar spade]

Do you normally log in with your phone number?

[candid cliff]

No

[polar spade]

Was the account originally set up through the app. Or did you use desktop?

[candid cliff]

Usually it stays logged in

[candid cliff]

It was created from my phone

[polar spade]

So you originally used an SSO or an authenticator portal, assuming Apple, but maybe Google, Facebook, Discord, MSN?

i.e. you clicked on one of the icons.

[candid cliff]

No clue, I downloaded the app, subscribed right away and never logged out again

[candid cliff]

I went back in using the google icon, my credits is there but my music isn’t

[polar spade]

Likely , as many do, including me in the beginning, we accidentally Signed-UP into a new account.

But wait... how many credits are showing?

[candid cliff]

Around 10 thousand

[polar spade]

When logged into this odd account,

1. does your history show any played songs.
2. did you have Premium before?
3. is your user/handle and avatar the same as you remember?

[candid cliff]

Everything is empty as if it were the first time I’m using the app

I’m not sure about the avatar I never paid attention to that

[polar spade]

Did it ask for your phone number to confirm youve clicked the correct authenticator icon?

[candid cliff]

No

I tried to use my phone number to login but it gave me an error every time

[polar spade]

OK. Which icon did you click to log in?

[candid cliff]

Google

But I don’t think that my original account was created using an email

[polar spade]

Google requires you associate one of your ggmail or registered emails with ggooggle. Thats how thry "authenticate" us.

So if you created with a phone number originally, that could be something to the creation of a new account by clicking Google.

What was the "error" given when using phone number. It may have just been a notice.

[polar spade]

And are you outside the USA?

[candid cliff]

No, I’m in Miami

The error said "something went wrong, try again later" and now it says that I tried to many times

[polar spade]

Miami is definitely another country. ( just kidding)

OK. So thats good. Your country code would not change. It would remain a +1

[polar spade]

Well, the "try again later" may be correct. Theyve been working through some micro-outages.

Have you tried through a desktop browser, on computer.

[candid cliff]

No, I will definitely do as soon as I get to the studio

[polar spade]

One quick attempt that's simple. Close the app. Click airplane mode on. Wait a minute. Then back off. Attempt a phone sign in.

[polar spade]

Granted these phone sign in simply have intermittant issues that are beyond our control. Waiting it out an hour and trying again later can be all thats needed. But making several attempts can block it from any current resolves. At least get the one hour wait time out of the way first before further attempts.

[tiny sierra]

Hello, for four days now I've been getting this error when clicking the create button: An error occurred.
A system error has occurred.
Account name: mmhqz23385

[candid cliff]

I’ll try that too thank you so much

[hollow plinth]

Can someone help me with trying to create a Memphis pooh shiesty type sound?

[neat nexus]

...I think I know some of those words.

Assuming this is a genre I simply haven't heard of. You will probably need to wait for someone who knows the sound or provide some examples.

[weak cedar]

this is solid. i like the negative prompt

[frozen dagger]

Dark modern Memphis trap with a cold, menacing atmosphere.
Slow to mid-tempo half-time groove around 135–145 BPM.
Heavy, punchy drums with dry snare hits and sparse hi-hat patterns.
Distorted, gritty 808 bass that follows the root notes with minimal slides.
Minimalist minor-key melody using dark piano, detuned bells, or eerie organ tones.
Raw street energy, tense and intimidating, stripped-down and aggressive.
Vocals are low, chesty, and authoritative with minimal autotune, very upfront and dry in the mix.
Overall vibe is cold, grim, and confident — pressure over polish.

not a refined prompt but a general idea.

[quiet solar]

It seems to be pretty good at steering the mix/mastering/production elements - or perhaps put another way, it's a good reminder why it can be useful to include those elements in style prompts.

It's also definitely been trained on the collective Suno folklore and voodoo of the internet, as it very confidently prompts for things I'm about 95% sure don't have any influence and suffers a bit in general from the "if I add lots and lots of complicated prompts the end result must surely be better!" fallacy. The majority of its lyrics prompts were straight up ignored.

I think people seeing "Hey, did you realise you can try prompting for all these different types of things" and being given the musical language for what they're asking for is a really helpful step in the learning curve for using Suno, but a tool like this isn't going to be a silver bullet for making everything sound better, or be more like the song people are trying to get out of their heads and into music.

[eternal nacelle]

I need help! I have a song which just needs a 3 second patch to replace a few words in the lyric. I have studio but I don't understand how to get the REPLACE function to regenerate just that little section. If I activate my microphone and sing the corrected lyric it does replace the section but it just puts my voice on there without the band. What I want it to do is use my voice just as a guide track to regenerate the lyric. I tried pasting my lyric into the Replace lyrics button and also into the lyrics box on the left. But its not doing what I need. Help!

[quiet solar]

@eternal nacelle Go to the version where you've sung over the section in your Library and Remaster it with "subtle" changes in the first instance. That should, hopefully, get the existing singers and instruments to continue over the part which you sang in.

If not, the fallback option is to Cover the entire song with low weirdness, low style influence and mid-to-high audio influence (e.g. 50%-85%). It might take a few tries to find the sweet spot, but that should have the same effect - using your singing as a guide which the 'band' plays in the final version.

If you're trying to make a really surgical change, where the rest of the song is completely unaffected and just those lyrics are replaced, there are other ways to do it using Studio, but it's a bit more of an involved process. Let me know if you want me to go over it. 😉

[eternal nacelle]

Thanks I'll try it. I've been asking Gemini for advice but realizing that for all its bravado, Gemini doesn't really know the SUNO UI.

Yes I am trying to make a surgical change. What are the other tricks in Studio?

[safe maple]

I created my account in Miami, and I'm on holiday in Montreal, there was outages I think twice yesterday, but for me, it was fixed, I've been creating since last evening. Have you tried another browser, or dumping cache for the last hour?

[quiet solar]

There is a way to do it, but it's more complex than just tinkering with the UI or prompts.

To explain; anything that Suno creates is always created again from scratch, so you'll always get new instrumentals and new vocals generated whatever approach you use. Sometimes the differences will be quite subtle, sometimes less so. Replace lyrics doesn't get around that, but is a quick and easy option for someone wanting to make a fix who isn't worried about minor changes to the instruments in the background.

If you really want to keep the instruments identical, you have to find a way to work with just the vocals. There are a few ways of doing this, but as you're a Premier user, what I'd personally suggest is this:

• Go to your original song in the Library and Get Stems, split it by instrumental and vocal.
• Extend your original song from somewhere after the section (e.g. Verse, Chorus) that you're looking to make lyric changes to.
• In Extend, put in the new version of the section that you're looking to repeat. Keep the style prompt identical unless there's something specific you're looking to change. Set Weirdness low (c.5%-15%), Style mid-low (15%-50%) and audio influence mid-high (50% - 85%). You may need to tweak these, particularly audio influence, to get the sweet spot.
• Extend as needed until you get vocals that you like and that fit quite closely to the original. Disregard the instrumentals.
• Get Stems from that Extension and split by instrumental and vocal.
• In Studio, start with the original instrumental stem and vocal stem and then import the new Extension vocal stem and use part of it to replace the original vocal stem in the section you're editing. For consistency of sound, you may find it easier to replace the entire section's vocals, not just the specific bit you're looking to change.

[safe maple]

It tells me it does, and I give hints with [instructions for instruments] and (attributes for lyrics) and it, for the most part follows then. Building memory on good interactions, corrective or otherwise is helpful. I've been getting awesome stuff

Was lead here:
https://github.com/RosCollective/suno-drivers-guide

Suno guide, very usefull

[quiet solar]

@eternal nacelle Even if you're going for a very surgical change, I'd suggest giving the Remaster/Cover routes a try first because

a) They're a lot faster and easier, and no real harm is done by trying
b) With high-ish audio influence, provided the song doesn't have lots of complex things all going on at once, Suno is pretty good at replicating the original closely, barring some slight adjustments to the underlying instruments or voice

The surgical route is a workaround for when you really, REALLY don't want to change a single thing about the instrumentation and are comfortable with audio editing to get round it.

If I haven't explained any of the steps clearly, just let me know.

[next peak]

Hey folks - New here! Premier/Studio user. Is there a way to have Suno generate a song without a bunch of "hiss" artifacts? My upload song recording has no such noise or hiss problem. Very clean upload. The intro is a simple dry drum loop with a warm synth chord that isn't busy at all, and Suno spits back a very hissy, noisy track from the jump. I tried asking this in Suno Studio chat but there doesn't seem to be much activity there. Does Suno generally not do well with full instrumentation uploads?

The prompt isn't a lot - "indie folk, indie pop, melancholy, cinematic textures, cinematic strings, upright piano" Are there some no-no keywords there?

[visual patio]

I dropped those styles into gemini and it says:

In Suno, the specific styles you mentioned are susceptible to "hiss," but for different reasons. While Suno has improved significantly with Version 4 and 4.5/5, "hiss" usually manifests as either a style-mimicry artifact (the AI trying to sound "vintage") or low-bitrate noise in quiet passages.

Here is how each of your styles typically behaves:

1. High Risk: Indie Folk & Upright Piano
   These are the most likely to contain hiss.

Upright Piano: Because "upright piano" is often associated with "intimate," "home-recorded," or "vintage" sounds in the training data, Suno often injects a floor of static or "felt noise" to mimic the sound of a real room or a physical piano mechanism.

Indie Folk: This style often triggers "acoustic tape recording" or "lo-fi" aesthetics. Suno may interpret this as needing a warm, analog hiss to feel authentic.

2. Moderate Risk: Melancholy & Cinematic Textures
   Melancholy: The risk here is quantization noise. Melancholy tracks are often very quiet. When an AI generation has long, quiet tail-ends or sparse arrangements, the "latent noise" (the static the AI uses to "dream" the music) can become audible as a metallic or "underwater" hiss.

Cinematic Textures: If the textures are "ambient" or "airy," you may hear a persistent high-frequency sheen.

3. Lowest Risk: Indie Pop & Cinematic Strings
   Indie Pop: These tracks are usually more "compressed" and "dense." The louder volume and busier frequency range tend to mask any background hiss.

Cinematic Strings: While strings can be quiet, Suno tends to treat "strings" with a higher-fidelity "hall reverb" profile, which is usually cleaner than the "room noise" assigned to an upright piano.

it suggested adding the following to try to get rid of the hiss: high fidelity, studio recording, clean, polished, 4k audio, noise-free

[next peak]

Thank you! I'll give those a go!

[visual patio]

Gemini is good for that kind of thing, also if you want to "Mimic" an artist it seems fairly accurate at giving you something that get's close without using the artists name.

[next peak]

Are you just asking about Suno prompts in regular Gemini Chat?

[visual patio]

yep

"in suno will any of these styles create hiss in the background? indie folk, indie pop, melancholy, cinematic textures, cinematic strings, upright piano"

[next peak]

Thanks again, this is really helpful

[rich ridge]

suggestions for prompts to get songs like this one: https://www.youtube.com/watch?v=kY4uAUZ0RrQ

[fresh knoll]

Use GPT to research it, but maybe underground rap with techno of some sort

[rich ridge]

yes did that well not chat gpt but asked grok but the most helpful was just looking at the artists page on spotify and seeing how it categorized his music and what playlist their ai added it to

"cosmic phonk" seems to get some good stuff

[fresh knoll]

Yeah. Used to be able to use avant-garde, but that reverted to the original style it's piano like. It's seems a build song not a genre to genorate

[rich ridge]

it seems to really like generic music lol

[fresh knoll]

Yeah it's not the same as it used to output. Avant-garde for a bit was like that in v3, but after it turned to the original style

[rich ridge]

have you tried any opensource models?

[fresh knoll]

No I just been self-taught mostly

[earnest agate]

mixed mode prompting is the way to go. Technical-Behavioral Hybrid Prompting (or Engineering-Descriptive Fusion)

Who said Suno can’t do emotion!! 🤖 🔥
https://suno.com/song/e338ca06-6d95-40c1-a13d-037c24590d01

[dapper storm]

????????

[rich ridge]

lol

[fresh knoll]

Just remove that word, and you will be fine

[wise kindle]

trying to understand what kind of style tag would require that particular word

is it a style of music I'm not aware of

[fresh knoll]

It's from gpt or something it's common output in style. It's actually a music term, but also several other meanings

[fresh knoll]

Yes

[polar spade]

Yea. Its like a reggae blues vibe. If I recall. Honestly, I think it has other usable genre or style terms, but alas...... skank is more fun to use, I suppose. Again, last I checked. It comes up about every four months.

[rich ridge]

lmaoooo i love you

can't say things main stream music talk about

[fiery bobcat]

Id love all male vocals to noylt h idk ALL SOUND THE SAME

[safe maple]

Dl it, edit to give a 3 minute sample (i could not upload a larger file, or it could not analyze longer than 3 minutes), re-encode mp3, upload to ask it to analyze the segment, ask it to give you the style minus the band or artist /human name

[safe maple]

I think that's what I've been doing, giving it a qwen analysis and reverse engineering lyrics
https://suno.com/s/C7EHDt0sdJgbbSTH

[rich ridge]

neat lol

[safe maple]

Cheers XD It's a labor of exploriation

[eternal nacelle]

Thanks again for all the suggestions. I'm trying them all one by one. But its a crap shoot no matter what since the track always gets redone. The first method - Remaster Subtle - still sounds too much like me, and if I crank it to Normal - then everyhing else gets changed too much. Cover also changes everything else too much. But I found a take that gets the passage right using the AI's voice. Subbing that in and then using Remaster -Subtle on that does the trick. Thanks!

[crimson pine]

Well my original message is gone, how ironic lol

[earnest agate]

Yup, 🔥 Sounds great

[agile plinth]

Why don't you share the style in your Suno songs?

That is the only thing worth sharing, no one wants to listen to your music. We make our own music

[hard beacon]

Well, well... you could say that in a nicer way 🙂

[earnest agate]

😃 i dont even like listing to half the music i create.

[livid crescent]

I dunno... I've gotten to where I'd rather listen the the stuff I make here than anything else

[quiet solar]

Whaddya mean you don't want to listen to my eighteen minute long Bulgarian deathcore ballad about toast and waffles?! 😂

Have to agree that coming to the prompt development channel, linking to a song and saying "This style of prompting is AWESOME. Whoa, whoa...no way am I sharing my style prompt..." isn't really within the spirit of the channel. 😉

[forest brook]

Hi,
I have a question , how I can make a clear prompt that suno is not mixing up the voices vocal and make it more clear on the song when I like to do audience responds. I tried following but the result is often an own interpretation of suno :) . Maybe anyone have a helpful advise what I'm doing wrong ?

[Male, Boy Group]
--- lyric ---
[ Female, Girl Group, large crowd, audience chants]
--- lyric respond call background audience --
[Male, Boy Group]
--- lyric ---
[ Female, Girl Group, large crowd, audience chants]
--- lyric respond call background audience--
[Male, Boy Group]
--- lyric ---

thank you

[earnest agate]

You guys are funny. OK, so how many layers are there to a Suno prompt, and what does each layer do? Just the Suno prompt? When you ask your LLM of choice to help you create a Suno prompt, what are you asking it to do? What is your priming prompt?

[neat nexus]

I'm assuming you've tried just this with the word duet in the style prompt?

I'd add some other terms for groups in the style prompt too, to try to emphasize the goal here. Maybe include call and response, audience participation and things like that to help you nudge it the right direction.

But you may need to just take the one that frustrates you the most, (you know the one that starts out almost perfectly?) and extend the track from right after the good part but before the bad continuation.

As an example:

[Male, Boy Group]
--- lyric ---

his part ends at 1:23

Extend from 1:23 and remove previous text starting with

[ Female, Girl Group, large crowd, audience chants]
--- lyric respond call background audience --

You might also go with changing the style prompt here to focus the vocals. The goal is to get it to do it once then you can try covering the full output and seeing if it retains the pattern afterward. Good luck!

[forest brook]

ok thank you . based on different threads and GPT I tried stuff like : (but not duet yet ) .

Boy Group Energy on Lead Lines,
Girl Group Vocal Hooks and Teasing Chants Throughout,

[neat nexus]

Use ChatGPT (and the others) but do not take their output as gospel. Use their output as inspiration but do not assume it actually knows anything about how Suno actually works. It doesn't. It's guessing. It may guess right sometimes though so it can be useful.

[forest brook]

sure. Their are two different AI's, but anywhere I have to start with my learn path to make errors in a success :) Thank you for your help .

[earnest agate]

have you tried something like this?

[forest brook]

from the idea ,yes

[safe maple]

Cheers!

[fringe pebble]

afternoon guys

[fast plume]

Good morning/afternoon all! I primarily use Suno to remix video game tracks from the retro era, what are some things I should do to make these remixes even better? I almost never mess with the sliders.

[polar spade]

Its a good read, but Im not sure what "audience does not carry lyrics" in comparison with Male and Female lead prompts for the same. I mean, when we can prompts audiences to sing through the whole chorus as much as we prompt for gender.

I think the over-all point of trying to force a genre as being difficult is relevant though.

So, exploring and experimenting is still warranted.

[frigid orbit]

Once getting one flavour of voice is easy, crowds, whispers, echos, glitched etc. should be relatively easy.

[pure panther]

Autobot censorship yes, refreshing lack of rules and strict mod control, also yes. I prefer the bot here to some crazy strict servers elsewhere.

[quick wedge]

It’s always been a +13 age server like it should be since Discord have it as well on their platform. I like the censorship where people can’t use the swear words.

[rich ridge]

[quick wedge]

Not everyone will like that sort of thing anyway

[safe maple]

Happy New years eve mate

[livid pewter]

I got the pro plan but I don’t have the pro features only have 50cr and don’t have v5 even though I paid for the plan

[frozen dagger]

if u just signed up. i've heard it can sometimes take few days to process.

[livid pewter]

No this is the second month

The first time it worked perfectly

And I can contact Suno

Can’t

[frozen dagger]

it renews on date you sign up. not 1st of the month. can see that info when you click your credits. maybe not time yet¿

[pure panther]

Yeah, and it reloads at the same time you got it. I had to wait until like 10pm, it was agony lol

[livid pewter]

No, it already charged me and the purchase already went through

It charged me yesterday

[pure panther]

Huh, probably best to contact support and throw a message here:
#archived-ticket-help

[pure panther]

Do you need the support email?

[livid pewter]

I already contacted them, but five days already passed, and they still didn’t contact me back

[pure panther]

We really can't help here, we're all just users. Try making a message here: #archived-ticket-help

[livid pewter]

Oh ok thanks

[polar spade]

From whhat I read , You bought a 1 year Pro plan. It activated the plan for 30 days then reverted back to Basic.

You bought the plan on November 28th?

[crimson pine]

Is remastered song actually better? My headphone is too bad to tell

I'm using subtle

[pure panther]

It depends. Sometimes it comes out better, sometimes it comes out worse.

[eternal nacelle]

Gemini thinks SUNO responds to all sorts of tags that I am skeptical of, but Gemini insists they have an effect. Obsure artists, tv shows, movies.

[pastel elm]

Anyone have advice on getting Suno to produce 2 distinct vocals consistently?

I've been putting together rap battles for funsies, and while I can get one character's voice right or at least close enough to what I want, Suno either runs the entire song in that voice or an entirely different 3rd voice I didn't want.

For example, this (https://suno.com/s/T3LChrSHR147vfm9) is one of songs where it ran with one voice the entire time.

The lyrics have the different characters, the verse, and the vocal style in brackets, and the style prompt gives further detail on the voice styles and rules for 2 distinct female vocals.

Does this just boil down to playing the RNG game with my credits hoping I get what I want, or is there some additonal things I can add/remove to have Suno be a bit more coorperative? Thanks ❤️

[frigid orbit]

try no [tags] or instructions in the lyric box and putting the "other vocals" encased in "quotations".

[quiet solar]

Two distinct vocals of the same gender are quite hard to do in a single song - you had the right idea giving the different registers, but it's very hit and miss.

If you're doing a song with two vocalists changing over frequently and not harmonising, you may be better off solving this by a mix of prompting and editing. Create a song where one singer performs the whole thing, then do an Extension of that song from most of the way through and repeat the entire song; this time prompting for the different singer to perform it completely.

Get Instrumental and Vocal stems from the original song and Singer 1, and the vocal stems from the Extension (Singer 2) and replace Singer 1 with Singer 2 as desired in Suno Studio or a DAW/audio editor of your choice.

PS. As a general point in prompting. Suno isn't a LLM and it only has a limited ability to interpret natural language. In your prompt, phrases like "...vocalists are female but must sound clearly different" are unlikely to have any effect. It also isn't really capable of factoring in conditions like "only" or "do not" - and it doesn't understand negatives. If you say "Do not blend voices", Suno will more often than not read "blend voices".

You're much better off telling it what you do want, and putting anything absolutely critical to exclude to go into the Exclude Styles box.

[jagged vector]

So how do you combat when the AI is completely ignoring the prompt, AND actually doing the opposite?
I guess yes really does mean no eh?

[weak cedar]

Keep trying by amending the prompt or exclusions

Lot of AI is really just experimentation

[grand wadi]

I find with the suno algo if u want it to respond - or carry out certain lyrics in a certain way its always best to put that particular action/request prompt in between { }

[sharp kettle]

i saw sumone mention duets. in my quick chat scan. it hit or miss but put this is song lyric helps
[chrous] [duet][ male voice 1: bartone. harmonizes with, male voice 2: tenor]

also prompt int the more ai decription

Production: Two male voice harmonize on verse 1 verse 3 and chorus, male voice one is baritone, male voice 2 is tenor....

its still gonna be hit or mis.. until we get a duet/ solo slider. suno tends to favor solo singers.... unles the genere is already filled with many sing pull models like [choir]

quick not you can do song names but i got a warning for tryinga tempo like Ice ice baby
Production: play in a style of ice ice baby. cant post artist but it kinda pulls from song names lol
pullled great in 4.5 asl and 5 for Kitty Kat songs, and i used burning love... didnt like the fact it pulled from the non famous version but hey i got 4 usable beats lol

[earnest agate]

Here are some examples, but it’s definitely hit and miss. Keeping Suno on track while switching between Voice 1 and Voice 2 isn’t easy. I know some people have gotten it right, but I’ve been trying to do it using only prompting, without using the editor.

https://suno.com/song/b4d7d83e-0d8d-4245-b8e6-557c8c571201
https://suno.com/song/1a621d5d-85dd-4915-9e59-7ef1fdb6f1bc

[hard beacon]

Happy New Year to the Community 🙂

[eternal nacelle]

We shouldn't have to guess about what syntax works in prompting. Is there a definitive guide from the developers?

[grand wadi]

not really sure about the guide..i am not aware of any currently

[earnest agate]

There is some good stuff in there. Always a good place to start
https://discord.com/channels/1069381916492562582/1252267444773326848

[eternal nacelle]

Gemini says Kanthric is 100% correct on the practical side of things, and he’s saving you a lot of wasted credits.

While Suno’s underlying architecture is technically a transformer (the same "brain" structure as an LLM), it’s trained on audio tokens, not conversational logic. It doesn't "reason" through your instructions; it associates keywords with sounds.

Here is the breakdown of why Kanthric’s advice is gold:

1. The "Pink Elephant" Problem (Negatives)
   AI models (including most image generators) are essentially associative engines. When you type "Do not use a flute," the strongest token the AI sees is "FLUTE." It doesn't have a logical "NOT" gate.

The Fix: Never tell it what not to do. Just flood the prompt with what you want. Instead of "No rock music," use [Style: Pure Jazz, Lounge, Acoustic].

Gemini says Curly Brackets are NOT recommended anywhere in the Suno wiki.

[hallow viper]

Is there a reference document that captures the syntax that can be used to maximize control over lyrics and style? Or has Suno really left that up to the community to figure out?

[grand wadi]

gemini has never used suno before tho lol

[eternal nacelle]

There's a Suno wiki - but its not official.

You're right Gemini just makes suggestions it doesn't know if its suggestions really work. But I'll tell you that after hours of experimentation with Gemini's help my outputs are a lot better than before I started following its suggestions.

[grand wadi]

Are you guys using V5? i think that also makes a big difference in the A Is ability to reason these situations out

[eternal nacelle]

I only use V5.

[grand wadi]

ok cool..honestly being a musician/audio engineeer etc before i found suno..discovering things about suno and using it is the best part about it for me

its an amazing tool

A I as a whole is so fascinating

[eternal nacelle]

I have some very complex projects and Gemini has done wonders to keep them from going off the rails.

[grand wadi]

Gemini is INSANE! I have used almost ever other of the type and nothing comes close to it

[hard beacon]

i have to say, that GEMINI gave me a lot of good infos in the past 😉

[plain pumice]

Anyone of you having issues with the Studio after starting, not many moments after it will begin to run sluggish and slow, and then the page will crash?
The error message given: Out of Memory.

[sharp kettle]

[verse 1][male vocals, baritone] [genre: hip hop]
sing this song but skip the space
[verse 2][female vocals, tenor] [genre: hip hop]
to combine the duet you go...
[chrous] [duet][ male voice 1: baritone. harmonizes with, female voice 2: tenor][harmony]
enter the harmonious song baaby

no real guees work you litterly just told the ai how to sing.
suno has a choir genre tag. it would be nice if the had the call&response music tag because there is alot of music video sample to pull from for the trainer its a popular music styel in 90s and 2000s music fields lol

[pure panther]

Sooo, I'm trying to add an instrument with studio, but it sounds disconnected from the beat and rhythm of the song. Is there a trick to getting it to follow the existing audio?

[hard beacon]

take a look to the official suno youtube page.. there is a video how to fix it (if i am right)

i think it was the first video where the show us the suno studio thing

[hallow viper]

I'm finding that in the Styles prompt the following format is working well:

Genre:
Mood:
Instruments:
Tempo:

After that I am adding the song structure that I generally use for lyrics - [Verse 1], [Pre-Chorus], [Chorus], etc. After each I am adding a brief description.

I plan at looking how that might interact with the structure of lyrics.

@frigid hearth Do you have ideas on how to use the Lyrics prompt to just produce vocals without specific words?

[earnest agate]

can you show a prompt example, please, because i cannot get suno to do anything useful using {}

[hard beacon]

{ ???? try [

[earnest agate]

yeah, [] work 100%, but i am curious about {}

[hard beacon]

ahh.. ok.. then i've missunderstood your message 🙂

well.. try to add "male vocals" in [ ] too 🙂

[earnest agate]

interesting, when i get a chance i will test like that and see what Suno comes up with

[cinder cargo]

heres how i do it in my lyrics :
[Verse 2 – Rap/Sexy Playful Switch]
[Bridge – Melodic/Delay Play]
[Pre-Chorus - layered harmonies, rising intensity]
[Bridge - half-time, intimate whispers building to full energy]

heres my profile , u can listen to some of the songs and see the results , tho i do clean up the lyrics from those before i publish the songs https://suno.com/@sheaperd

you can also see the entire prompts , that structure seems to bring me the best results compared to what i see being used

[plain pumice]

With regards to this:

The error message given: Out of Memory.```
To note, when you pay attention to the taskmanager for the specific tab - Suno | AI Music, you can see it steadily rises despite nothing is going on in the Studio. This means there is a Memory leak.

[quiet solar]

There isn't an official source, because that's not quite how prompts work. They're not commands in a programming language designed to produce (y) result every time you put in (x) input, but a way of requesting that Suno take into account something that you ask for when generating either the entire song (style box), or a particular section/line of the song (lyrics box, and to some extent style box).

Anything you put within [ ] in the lyrics box, Suno will typically recognise as being a prompt. That doesn't mean it can do what you're asking for, or that it will do what you're asking for even if it can, but if the prompt is picked up and given importance, then Suno will try to look up what you've prompted for in its musical training library and factor that in.

In fact, even saying "anything in square brackets is a prompt" is a bit of a simplification, because people have found other ways to get to distinguish prompts within the lyrics that don't use square brackets, since Suno can somewhat evaluate context - I'm just framing it like that for the purposes of keeping this as simple as possible. 😉

As such, there isn't a readily available official guide aside from "put things in, try it out and see what happens". If there are prompts that the devs have particularly trained Suno against, we don't have a list - but musical terminology tends to work more effectively as a general rule, as do prompts impacting the general style or tone of the song rather than prompts focusing on something particularly detailed, like a specific type of intrustment or chord pattern.

[quiet solar]

Have you tried anything along the lines of [ad lib vocalisation] in the lyrics box, or using "oooh-aaahhh" (etc.) as lyrics?

[safe maple]

Have you seen this yet?
Was lead here:
https://github.com/RosCollective/suno-drivers-guide

Suno guide

[quiet solar]

Switching between Voice 1 and Voice 2 goes beyond what the model is capable of, because it doesn't necessarily attempt to factor in every single prompt it comes across. Any song in which you require Suno to do a specific thing at a specific time is incredibly difficult as a result.

Having a song use two voices is 100% possible, and having those voices typically switch over based on where you put prompts is 100% possible. Anything more precise than that is fundamentally down to luck if you're just relying on prompting.

You can try to tilt those odds a little more in your favour in the prompting you use, but all of the best advice on how to do duets reliably in Suno either involves making use of an editor (be that Studio or an external one), or repeat use of Extend to 'bank' bits of the song that went correctly as you're going along, which is a glorified way of tilting the odds in your favour. There is no way to make it work reliably through prompting alone in the versions to date.

[quiet solar]

I believe Suno can recognise the term "call and response"; past testing by myself and others suggests it does make it more likely that the two voices will alternate line by line when you use it. Whether that's helpful or not depends what you're trying to do in the context of the duet as a whole.

[earnest burrow]

How do I create an iPhone call intro for a song, with no music at first, just the iPhone ring and a female voice saying ‘Hello, how are you

[grand wadi]

I literally have a whole "call and response" male and female track on my suno profile

Sometimes just re covering the same song over a few times can tighten results

Along with the proper prompt structure

[earnest agate]

"n fact, even saying "anything in square brackets is a prompt" is a bit of a simplification, because people have found other ways to get to distinguish prompts within the lyrics that don't use square brackets, since Suno can somewhat evaluate context - I'm just framing it like that for the purposes of keeping this as simple as possible." yeah, i have always thought that as i had some very cool songs back in v4 (i think it was) that were giving me amazing results. Some nice songs, but so random. So yeah, i also think in some way "Suno can somewhat evaluate context " but i have never been able to pin done how or when it doing it.

[hallow viper]

I haven't, no. I will experiment with that. Appreciate the suggestion!

[earnest agate]

This kinda work, but as said above, Suno does not explicitly adhere to Tags or "process" them in order

[plush field]

just a side note, negatives are perfectly fine if they are in proper place (Exclude style) and in a proper form (without any use of terms like "no", "without"...)

[plush field]

as Kanthric told you, you can use oooh-aaahhh, etc...
but i would also suggest putting those in parenthesis, so they be considered as backvocals

so
(Oh-oh-oh)

instead of
Oh-oh-oh

[toxic pasture]

Sorry if this has been asked or mentioned, but is anyone else having audio issues with Suno today? Every song has high/low gain issues. Loud and soft alternating...

[flint zephyr]

how do you guys get around the Standard Algorithm, seems as if the key or tonal center is the same across every genre I work in, and despite prompts directing it to modulate down it seems it ignores directives Im using [xxx] should i maybe use a different bracket

Audio issues Galore today V5 is trash anyway, V 4.5 is stable in my box at least

[plush field]

if you use the same song name or similar enough, suno will "continue"

#📚┃suno-school message

[earnest agate]

TEST 1: Stacked Tags Method (Hip-Hop Duet): This actually seems to work. Still need to be tweaked a bit or fun a few more generations.
https://suno.com/song/c9600384-e89b-4f85-889c-0b927e762605

[frozen dagger]

Parenthesis* sorry. not trying to be grammar police. just... u know already.

[plush field]

english is (obviously) not my first language, for me brackets are those which are not square brackets 🙃 but, yes that is the proper term

[frozen dagger]

ah i see. just trying to make sure new people not get confused. parenthesis ( ) are for backing vocals. Brackets [ ] are for non-lyric insstruction.
🙃

[plush field]

i agree 100% with you, so i fixed the sentence

[earnest agate]

Stacked Tag Technique Detailed Explanation
https://discord.com/channels/1069381916492562582/1452031031032025109

[quiet solar]

I've never personally found any noticeable difference between stacked tagging and identical unstacked tagging, either in terms of effect or likelihood of Suno paying attention to the instruction.

All I would say is it's not a great idea to use this syntax:

[Verse 1]
(Male voice, spoken delivery)

Because () brackets are typically used in lyric notation to indicate a backing singer or echo of the main singer's line, and Suno picks that up. And most people don't usually consider it a success when Suno sings their prompts. 😉

[quiet solar]

If you're looking to set the key for the song as a whole, you're probably better off putting that in the overall style prompt rather than in lyric prompts, which would be better suited for attempting a key change mid-song.

[ your prompt here ] is the solidly reliable way to include a prompt in the lyrics section. It's not the brackets getting in your way, but might be how you're asking for it to make a change within the prompt itself. Feel free to share if you want direct feedback.

[lone owl]

Guys it does BPM affect the results? Why always slow bpm result over use saxophone, first time well is fun looks like kenny g cooking in his studio, but i want only classic synthesizer. I feel sick and tired of this suno because doesn't follow style instruction or lyrics(the chords and melody should be exist). Any idea? Im free user.

[quiet solar]

@lone owl To a certain extent, everything affects everything else. If you prompt for slow BPM, Suno will pull in characteristics of other slow BPM songs - particularly from the genre you're specifying. If that's introducing specific things you don't want, like saxophones, you can try adding 'saxophone' to the Exclude Styles list under advanced options.

In terms of following or not following style prompts, it's best to think of all prompts as being suggestions to Suno rather than instructions - that's how probability models work. You can't 'instruct' Suno to do something, but you can tell it the things that you want and let it cook. As a free user some of your options there are more limited, but you should still be able to steer the song the way you want it to go - it might just take a few more tries when you're only able to use the default slider settings.

If there's anything specific you're trying to do / avoid doing, feel free to ask - people here are generally pretty helpful. 🙂

[nimble trail]

I find that using the audio upload MAJORLY helps with getting the bpm right.

[lone owl]

Oh thank you, anyway
Got some suggest for making 80s, 90s synth slow pace instrument?. Well cause im free user i cant use some excluded style anyway.

[nimble trail]

I was trying to get a sea shanty, and it wasnt getting the bpm right, so I sang the first few words in the bpm I wanted it, and had it cover my audio for the song

and from then on it got the BPM perfect

[lone owl]

I do some too, i put (where do broken hearts go - whitney houston) and E.piano 1 DX7 into but in suno turn into saxophone melody😂😭😩.

[nimble trail]

I also, when trying to get minute changes, edit some of the words

making a line shorter or longer can add tiny yet noticible changes

But for me at least I most of the time just let suno do its thing and just go from there, I worry more about the lyrics then the flow

[nimble trail]

https://suno.com/s/reipOYh54FRJyeQN

[earnest agate]

https://discord.com/channels/1069381916492562582/1452031031032025109 is the blue print to Suno ai. The song aint perfect and the lyrics are not great, but my word never got this out of Suno before
https://suno.com/song/20413213-a64e-472e-b3f2-f5b5b5862e6c
https://suno.com/song/b8ebff56-e7a1-41e1-b69c-0771d462a484

[tiny kayak]

[Clear Memory] works as a command right in a prompt to wipe any old artefacts replication?

[wise kindle]

Some people have mentioned [Reset Memory] [Clear Memory], though I haven't tested it myself so I can't say for certain it works

[steady crane]

Not really correct on certain stuff there such as instructions. I'm a professional paid AI engineer. But don't really want to go into jargon rambles. 😅 but you'll learn as you go.

[steady crane]

Lil look of me squeezing every ounce out of V4.

https://suno.com/s/wNTASrLSs0sg8D3o

[spiral aurora]

What's the best way to indicate AI stuff?

[Verse 2 | instrumental decreases]
blah blah blah
epic lyrics

Or

[Verse 2]
[(Instrumental decreases)]
blah blah blah
epic lyrics

Or another way? Thanks in advance!

[quiet solar]

On a technical level, it's an oversimplification for sure.

But in general I'm aiming to explain in more human terms what people will experience, and a more helpful way to look at Suno as a tool than "There's a magic prompt which will do exactly what I want, if only I knew the syntax or had the secret codebook, etc."

[crisp sparrow]

Has anyone had any success with adding a bass guitar solo to a song? I'm trying to do just that at the moment, but it just gives me a lead guitar every time. Really not sure what other prompt/technique could possibly help me in this situation.

[quiet solar]

That's a fantastic sound, and a good working example of the limitations of prompting; the song is at a much faster BPM than requested, it's not switching singers in the specified places, it skips some lyrics to keep cadence, disregards the instruction to whisper lyrics, comes in with the build in the bridge too early, etc.

Yet there's absolutely nothing wrong with any of your prompts.

And quite honestly, none of that matters if you're happy with the end result - it sounds great, even if it's not precisely what you had in mind on hitting Create. And everything on that list would be fixable through a combination of tactical regenerating and editing if you really wanted to keep it tight to the original vision.

[weak cedar]

You can try using [bridge] and then describe the vibe using more than just bass guitar (low frequency only) and experiment that way, or uploaded a bass guitar solo/isolated track and then to extend on it then rearrange in editor for the rest but it's tricky

I haven't had much success with solos tbh I try to do drum break solos sometimes and it just makes a bridge

@crisp sparrow

[quiet solar]

I've also had headaches along the same lines.

In the end I took the route of vocalising a basic idea for the melody, slapped that in the relevant place with an editor and had Suno cover the edited version with the style prompt specifying that there was a bass guitar and the section tagged as a bass guitar solo. That worked passably well.

[crisp sparrow]

@weak cedar Yeah I've never had any luck with drum solos either.

@quiet solar That's not a bad idea, although that may be quite difficult for me to try and create a section where the bass solo would need to go without knowing what it would be like. I've never been good at coming up with melodies - that's why I'm using Suno in the first place! 😆

[quiet solar]

@crisp sparrow Yeah, I hear you. A common-ish workflow for me would be to Extend just before the point where the bass solo isn't coming in, then rewrite a large chunk of the style prompt and the lyric prompting to focus specifically on getting Suno to make a bass guitar solo. Then once you (finally) have it, edit it into the original song using Suno Studio or an external editor.

But that depends on your comfort level with audio editing, and whether you feel it's worth the time and effort to try to work around areas where Suno struggles a bit to give you exactly what you want.

[frigid orbit]

I’d just do [Pause] ahead of the lyrics

Why are you spamming?

[lone owl]

Well for me trying created 1978-1982 synthesizer lead but ended up overused saxophone entire audio🤔🤣😂🙂🙃😩😭🥀💔

[earnest agate]

there is some interesting stuff there, thx for sharing

[sand hollow]

Somehow a straight forward prompt worked better than all the other ones i’ve tried

“While keeping the original song, add hip hop drill and trap into it”

[polar spade]

It looks interesting, but In practical terms, the 10 users only that saw them probably arent going to click on every one of those to read . . . and in 10 minutes, it's scrolled away to oblivion.

Why not just combine in a 20 page document and post in workflow channel https://discord.com/channels/1069381916492562582/1252267444773326848 for everyone, including me, to first read, then link to when new users need help and prompt suggestions, then discuss as well .

[sand hollow]

Though only sometimes it work, and with certain song genres it doesn’t. In this instance i’m on the mobile app so I was using the default “cover” and i believe the version is 4.5+

It’s pretty good with creating video game music

[alpine ore]

@final hazel Hi, can language "Moroccan Arabic (Darija) used in song"performed as a native Moroccan speaker of Darija?

[crisp sparrow]

Well thank you @quiet solar and @weak cedar for your input on the whole bass solo thing. What I've ended up with is definitely not a bass solo, but eh I guess it's different enough. 😆

And this is all just for part 1 of the project - I thought this would be the easy part! Lord only knows how many credits I'm about to burn through for part 2...

[steady crane]

Singer control isn't possible in current generation models. Was just experimental instructions you were looking at. I'm a professional paid AI engineer not hobbyist. To get tighter control we need a new model.

I got exactly what I needed to get out of it. The voice control was more or less flavor that isn't directly possible yet. Duets require something truly multi agent. Right now it's just one agent running 2 masks.

Speaking of Duet control.

Again. I'm paid for R&D. I know what I'm talking about. I'm not a casual user. Professional.

Lot of misinformation and misunderstanding about how this stuff works. 😅

But I'm not here to teach you AI or how the models work. That would require a lot of college type sit downs and lectures 🙃. Plus I do believe you learn these a bit better through your casual experimenting. I'm at most around to occasionally correct misinformation or occasionally troubleshoot.

But generally keep a low profile. Less intimidation risk rather than do tech jargon.

[steady crane]

But this type of model is a lot more simplistic than the stuff I'm used to working with. Limitations exist in particular control vectors. But very clear structured instructions get the desired output. Close to it as you can anyway on probability. Anyhow, gotta head off now.

[quiet solar]

We spoke about the limitations of duets and lack of a multi-agent model in this channel before, if you recall.

I'm sure no-one is doubting your professional credentials, and there's definitely a lot of misinformation and misinterpretation floating around. However, this is a channel where we get everyone from professional musicians using multiple tools and trying to fine-tune really complicated things that stretch the model past breaking point, through to people who are complete novices to anything GenAI.

Pitching advice at people at the level at which they're ready to engage with it is intended to be helpful, even if it's not a strictly technically accurate description of how the model works under the hood.

[pure panther]

One of the things I've learned about myself by using Suno. I really don't like the Theremin... Trying to negative prompt it into oblivion. I've learned it's also called, a synthesizer using a portamento (glide) effect, Electro-Theremin and Tannerin.

[steady crane]

As I said. Not here to lecture or really correct people often. My profession is essentially in model engineering/R&D. Knowing how they work under the hood.

Suno is a rather simple model. Not as complex as stuff I'm usually working with. 😅 In what it does and all that.

[pure panther]

Simple is probably good in this case. Airplane rule. Also, you don't eat steak at a spaghetti house.

[steady crane]

Depends on how they set it up in their infrastructure. Couldn't give you a 100% factual answer because I'm not currently in contract with Suno to see what they did.

Freelancer.

[pure panther]

Idk, about that. Who's reinforcing? And based on what? People have wildly varied tastes. That might work on a genre to genre model basis, but with it all mashed together I doubt there would be a very solid consensus.

[steady crane]

You'd be surprised honestly. AI is not simple. That's why my career area exists. But like I said continue having fun and experimenting. But know: No. It's not a simple thing especially once you get into LLMs.

Suno isn't a full LLM. Like all AI it shares that generation stuff and etc but it doesn't work the same. But it's why I call it technically more simple.

[pure panther]

Trigger warning

[earnest agate]

i’m not an AI engineer and I don’t have a lot of music knowledge either, although I am learning far more than I ever thought I would need to just by playing with Suno. You might be surprised at how complex a prompt Suno can work with. Sure, there are a few small misses in the generation, but it can produce a whole songs. It really is amazing what it can do.

I’ve come to Suno from a very different angle. I came here for the songwriting, so in many ways I’ve worked backward toward actually building a song. Because that’s what you’re doing here, you’re building something in layers. Well, that’s how I see it. my system is overly complicated, mad-hatter stuff, but it works, and it’s getting better and better after tuning each layer. My prompts run through, somthing like a 16 engine piple line, bulding the layers, like building a house.

[polar spade]

Its a nice compliment when the developers show appreciation to the common user who is finding out how to do things which the developers hadn't thought possible. So, knowledge and theory can limit creativity if we dont explore the impossible as well.

[pure panther]

Yusss, think outside the box.

Break the "rules"

[polar spade]

Its fun when we look at a kid do something amazing and catch us thinking, "Hmm, no one is supposed to be able to do that." And then we kick ourselves.

[pure panther]

Innovation = The introduction of something new.
You can't get that by following "rules"

[frozen dagger]

because no one told the kid it isn't possible. dream on.

[steady crane]

Some people here are still in theory or see it as black box. I'd break a of few those assumptions/theories on a few corrections so I'll just let you fellows figure it out. Lol.

[steady crane]

Course I need to stay lower profile anyways cause this crap triggers my loopy machine wired wavelength. Tend to repeat sentences and thought patterns including my own profession stance. 💀

[plush field]

oversimplified explanation of LLM that i like to use to explain to people who are not in IT is "autocomplete on steroids" 🙈

and for generative AI I like to use "imagine a "snow" from analog TV era, and then imagine that AI is trying to figure out how to make an image from that snow" 🙃

but i have no idea how audio models work, not even roughly

any suggestion for a oversimplified explanation for non IT people?

[steady crane]

Music generations share the fact that their probability machines. But Suno is not a LLM or advanced as one.

My best description is their generation buddies that eat prompts and spit out probable rhythms to what you instructed. Think of it like a Music slot machine. But better precision lever pulls gets closer to jackpot

The more structured it it is the more it'll understand. It doesn't understand "Sad song" alone that produces sloppy mush

[earnest agate]

Yeah, don’t get me wrong, Suno is no black box. There are lots of different ways to get something really nice out of Suno. Finding Suno’s “creativity” is what I’m looking for. I don’t like flat-sounding or two-dimensional music, and if you just leave it all to Suno, I find it tends to create that—even though it’s capable of a lot more. The model can have soul if you pull it out of it. I’ve tried many different prompting techniques, and each one produces something different. Some are really terrible thou. It a journey and that is why we keep putting coins in the slot machine 🙂

[plush field]

so, something like producing few different images with SD, with different prompts for each of them? where model first split each part to "generate" lower level prompts which it execute?

[steady crane]

You have to give it a BPM and etc. This gives the slot machine more direction if that helps makes sense of it

150 bpm, 85 bpm

[plush field]

that part is clear, but i am more interested what happes after tokenization

it tries to generate parts and than make a complete song?

so it goes trough prompt and says - ok i will use this for bass, this for vocals...

[steady crane]

Yeah it'll eat what's in there for example. I made it produce robot gears and grinding metal sounds by adding a explicit prompt layer with instruments that can generate it

Sounded like an terminator industrial thing

[plush field]

any idea if it has a different model for each part? for example model for bass, model for drums, and then something that mix them?

[steady crane]

I can't be 100% clear because I'm not in Suno's system and how they've infrastructured their model. Like how they got their V5 different from the V4 in detail

I know I've pushed V4 to sound like V5 just about.

[plush field]

i as far as i understand there can only be one of two - separate model for each thing or one mixed model that does all

if you consider stems it is more realistic it is separate model for each part, with "a mixer" model that combine those things

other possibility is that they use something like MDX23 to make stems from a finished track..

i will need to find some time to read more about audio generative AI

[steady crane]

You can hear some advanced noises in this one I did.

https://suno.com/s/YUt42XvuHibZ9wZX

Was coaxing V4 to make mechanical noises.

Ran another one with actual marching effects but the instrumental flexed over the speaker and made them sound pitiful in comparison. Poor AI singer couldn't match the fire instruments 🤣

[charred trout]

Hii

[earnest agate]

This should wake you up bob!!
https://suno.com/song/bd84fcf7-e42f-4ce7-884f-74d0b8fd9dc0

[hidden acorn]

a lot depends on which prompts you write. most only use words, and you can hear the differenece imediatly as music creator.

[steady crane]

Yes those are generally misunderstanding AI. Some prompts are architecturally incorrect. Though not many want to hear this. I won't specify as to not intimidate. But yes some prompt methods are "bad." And not using the system correctly. There is no such thing as all styles count they do not.

Some are less horrible than others but I have seen some incorrect methods. I just don't point at it.

[hidden acorn]

yea it needs a lot more then just words. I posted a short workflow example in the workflows section a few minutes ago, showing how I structure prompts for control and clarity. one thing is for sure Suno ist by far the best on the Market, i mad tests by others and they completely missed the mark. i emailed them and got no answer .

[steady crane]

I generally get basic rhythm with these 2 instructions.

GENRE: Electropop + hip hop influenced beats
TEMPO: 96 BPM • KEY: F Major

Tiny fraction of the thing but yeah.

Almost made nightcore once. 🤣

Wackest experiment was me getting terminator theme like drums to fire. Sounded about right but my instrumentals drowned the singer. They were too good for the speaker. Instruments were fire though.

Didn't publish though because the singer part made it sound meh

[hidden acorn]

well the ai can still produce a good track, i heared a lot of ai generated music on youtube, but many high-view tracks still show issues like vocal artifacts, limited dynamics, or lack of spatial depth fx-hall.

[steady crane]

I can smooth it into coherence. But there's always an edge case. Rock is inherently unstable

[hidden acorn]

that hurts, were ur publishing?

[hidden acorn]

exactly, thats why numbers are very important for the ai and you can always start and build a new track from it with the right tools, knowing very good how it will sound.

https://suno.com/s/hR32kFXHDdEvlsVC

warehouse raw techno from me

[steady crane]

Trying to locate the track. It's lost in so much sludge. I generated random crap playing with instruments before doing something serious with them

Some sound great instrumental wise just my singer was inherently drifty. Because I wasn't really tuning them to match

Not really listenable in normal sense but tech babble wise their fun to listen to

[steady crane]

Found it. Opens with almost a tribal war drum spam.

[hidden acorn]

I know that chaos very well — tens of thousands of generated tracks, lol.
You really need a structure that actually fits the song, plus tempo or at least key information and basic level guidance, so things don’t run out of control.
Otherwise you end up fixing everything afterward, which is exhausting and extremely time-consuming.

[steady crane]

Oh it was well structured. The singer no. I didn't patch him

[hidden acorn]

yea they do like it, its not my genre but, honestly said, as more as you create also different genres you start to like them too

[steady crane]

I was playing with instruments. They were great.

[hidden acorn]

its like your baby the song 😂

[steady crane]

Some good sounding ones were me dabbling into pop and electropop

[hidden acorn]

its our main genre what you listend since 4 ever i guess ?

[steady crane]

Not many followers here but to be honest they'd be fed up with the engineer spamming remastered experiments 🤣

I'm more known for my chat bots in hobby spaces

[steady crane]

I favor rock and pop/punk. Usually what I generate.

[hidden acorn]

chat bots do you train them ?

[steady crane]

Those RP bots. As a freelancer I'm mostly hired for poking crap, R&D. Sometimes hobby folks pay me for my AI art because it's not using usual prompting techniques.

[quiet solar]

As someone else who mostly creates in the pop/punk sphere, I'd be fascinated to see some of your outputs and the prompts behind it. I see you don't typically publish them on Suno (not that I can talk), but if you'd be willing to share a link I'd be obliged.

[hidden acorn]

those genres are great i also love them, when i listen music on my own then rock or punkr rock too a bit pop genres and mixed

[steady crane]

I share some decent songs on here but they are limited in variety. Haven't been around the app too long.

[hidden acorn]

ok was nice to meet you iam out again from the server. take care. and happy creating

[steady crane]

Will see if I can find one that's technically decent enough to send here

https://suno.com/s/WbhM0rUNpTqityTZ

[quiet solar]

Cheers

[steady crane]

My main win here is they fire my chorus the same way every time

[pure panther]

I stopped thinking of it as a science when my hubby got one of the best songs had gotten from his whole time using Suno after the used this simple prompt:
make this song drive across the country and move in with your internet girlfriend crazy

Sure, maybe there's a science to it, but it's certainly not the only way to get great results.

[steady crane]

This is incorrect but I can't really show you without taking you into college lecture territory. 😭

[pure panther]

Incorrect sounded really good

[steady crane]

As a professional who works in the field. There absolutely is more things to AI.

AI is probability yes some poor scaffolding will sometimes still produce something. But they aren't going to flex or show what the model is really capable of. That's the harsh truth.

But most casuals aren't trying to do that. Their doing fun prompts. Which is valid

[pure panther]

I keep reading articles citing professionals that say even the creators don't understand all the nuances of how LLM's work (I know Suno isn't an LLM) I can't help but feel like there's some over simplification and/or underestimating happening here.

[steady crane]

I work in lab field of AI. It's difficult to really go into jargon. But I know a bit of the black box as I have to do R&D LLM.

[pure panther]

That's really not saying anything.

[steady crane]

That's all I can really share.

[pure panther]

Playing it safe I see tee hee

[sleek stump]

Thats likely the NDA/employment requirement 😉

[pure panther]

Non statements can't be refuted. Also, I should say, since interwebs doesn't convey tone. I feel like this is a delightful light hearted conversation.

[sick peak]

https://tenor.com/view/popcorn-guy-relaxing-popcorm-popcorn-gif-26404903

[pure panther]

Can I have some too?

[sick peak]

https://tenor.com/view/bumby-wool-yoshioka-popcorn-offering-gif-25857539

[sleek stump]

I'd rather have Pizza 😄

[pure panther]

That also sounds amazing... what about pizza with popcorn on top?

I need to eat.

[steady crane]

Basically the best I can say is: Freelancer that does professional work with AI. Sometimes models, sometimes hobbyists pay me for crap. I can't go into every detail. Not all because of professional restraint but these are hobby fun bubbles. Last thing they want is some engineer talking above them with weird jargon 🤣

[sleek stump]

Pizza grease in place of butter? yes please!

[pure panther]

I love being talked down to with jargon. Please proceed.

[steady crane]

No. I hate being seen as superior sounding 😅💀

That's what it comes across as lol. "Me smart. Eat Tech babble" Bonking you with text book

[pure panther]

Uhhh, too late? 😂 Fr though, you're cool with me. But to be real with you, repeating your credentials comes off that way. I used to do it myself 😛

[steady crane]

I just can't stop myself. It's how my brain is wired 💀

I'm loopy. Very loopy

I have native machine fluency

It's not repeating out of ego it's my brain running self diagnostics

[sick peak]

https://tenor.com/view/fairly-odd-parents-doug-dimmadome-you-lost-me-there-i-dont-know-confused-gif-25462529

[steady crane]

It's why I'll usually just run that through an AI when I cant talk to people because they'll get annoyed by me. I repeatedly loop thought patterns.

[sick peak]

p.s. I like jargon, helps me learn... thanks for the quick detour. Not annoyed at all... I am going back to writing my prompts.

[steady crane]

Just have to clarify to some because it sometimes my loops are misunderstood lol. I.e me repeating a profession or some other thought

[sick peak]

I was actually intrigued by the conversation, first time posts for me in the server.

[steady crane]

Anyhow I probably need more tracks here eventually. Literally half my content is me remastering like 5 songs 💀

Too lazy to get new lyrics. 💀

[sick peak]

I feel like this is all I do... write, rewrite, scrub, rewrite, and so on...

[frozen dagger]

That's how it starts. Then your like @sick peak and me and the rest of us.

[sleek stump]

Writing your own lyrics is a whole nother' level of satisfaction, I highly recommend it. Even if its something you use an LLM to refine.

[quiet solar]

It also makes your song sound a lot, LOT less like every other song that everyone else makes. I'd also heartily recommend it.

[sleek stump]

Yes, so much this!

[quiet solar]

I really don't need to hear any more untold stories about neon stars and wonky metaphors. 😉

[sleek stump]

Its also fun to write dis songs abotu AI norms lol

[sick peak]

back to the lab...

[woeful bone]

[steady crane]

I've occasionally dabbled in some new songs in the past. May eventually finish them for suno. We see. 😅

Currently trying to finish a new rp bot for my players in some other platform.

[steady crane]

One was some fan made song for FNAF Roxy but never really fleshed it out

[sharp kettle]

thanks i didnt know its had call and response tag programe i just used duet tag
set up in lytic and repeated itt in production orders

[pure panther]

Same... I write them, then have to hear them in song form a bagillion times before they're "right", then a day later I hear something that could be worded better.

[sick peak]

Did you ever get a duet vocal to work right? I have not tried but interested in learning how.

[pure panther]

Yes actually, here:
https://suno.com/s/nycUv4wCc6PEoo5p

I made a male version, then a female version, then I roughly spliced them together in audacity, then reimported the audio into Suno and used it as a cover. There was more to it, but that's the meat of it.

[steady crane]

Its not fully possible on current models. Not full control anyway

Semi maybe but not full direct 100% control on making singers fire hyper specific lines

Can you make a general decent sounding duet? Yes just dont expect full control over it. Model will fight you and you’ll end up with alien dialect 😅

If trying to run the duet through Suno instructions.

[sleek stump]

Actually, I have have not had any issues getting duets to work as I expect. Most important part of them is to describe in detail what you expect from the duet in the styles window, then call them out in meta tags. I think my rate is 90% success.

[sick peak]

I just tried with vocal instructions within the lyrics, enclosed with [instruction], between the lyrics I want and it worked out very well.

[pure panther]

Call and response duets can work well, but if you want the man to sing the first 2 verses and the 1st chorus and the woman to sing the next 2 verses and the second chorus, that method doesn't work. At least it didn't for me.

[sick peak]

Example used in the lyrics..
[Vocal: Female Lead]
[Vocal: Male Lead]
[Vocal: Male & Female Harmony]

Not sure this would work well with persona... but a raw first pass seemed to work well.

[quiet solar]

Yeah, 100% this. Only use 'call and response' if you want the singers to alternate one line each, or it'll cause you as many headaches as it solves.

[steady crane]

It's still one of the trickiest generations. Primarily due to it's not multi agents. It's 1 agent playing 2 masks.

Can still sound good but very difficult to sound exactly right

[eternal nacelle]

Yeah and tagging [Male Vocal] [Female Vocal] doesn't work if the AI has other ideas.

[sleek stump]

True, when multi-modal gets to be a thing, duets are going to be very easy.

[quiet solar]

Getting two separate voices to sing in the same song is very possible.

Controlling them to the point where you can reliably specify who sings what just using prompting isn't possible. You can give Suno a steer, and make it more likely to do what you're asking rather than not, but unless you are very, very lucky there will be multiple mistakes.

It's still possible to get the end result you want, but it becomes a mix of workarounds or editing as well as prompting alone.

[steady crane]

When multi agents are a thing yeah. Much less glitches.

At least in getting duets to fire correctly

[pure panther]

And then you have to hope that it doesn't sound like garbo when/if it finally lands on the combo you want.

[steady crane]

The full control is what I essentially call tech limits of today. Multiple of the songs will misfire. Maybe 1 decent one gets through but luck based

[quiet solar]

@pure panther Haha. All too true.

[sick peak]

500 credits later

[steady crane]

That's basically what it'd become. Credit grinding to get it to fire even semi right 🤣

[sick peak]

It is interesting, the persona of male vocals was honored but the instructions of male/female did not land, not even close, still sounds great but not what I wanted 🙁

[steady crane]

My one duet I published isn't reading my full control attempts I did try but it does sound decent.

[quiet solar]

Which is why attempting it through prompting alone in a single song is a swift route to frustration.

Using multiple extends to build the song (optionally switching vocal personas each time a singer changes); creating two full versions and splicing (as @pure panther mentioned); or creating a master version and doing tactical splicing to fix it, are pretty much the best practice methods of creating duets in the current versions.

Each have their pros and cons, and only the first one can be done entirely within Suno and without editing, and it's a painful process if the singers change over a lot.

[pure panther]

And pricey

[steady crane]

Essentially. Also when I say impossible I do literally mean through using AI and not editing like splicing or using other tools.

AI can't make duets yet. On their own 🤣

Maybe once multi agent comes along

[earnest agate]

Here is my doc on Duets, mix of a bunch of techiques. Very hit and miss, but if it help give this a try.

[pure panther]

Welp, time to write some lyrics.

[steady crane]

[sharp kettle]

yes in my 3 tests
(didnt like the song way with my matched vocals so i delete but
step1 Lyric tag(edited to fit in call and response tag, a ad some spaces if you want it to do the song right.

[call and response}[production: female voice calls out to male voice]
[Verse 1}[vocal: male][call]
calling out to my babe
where are you now
call out to me
[chorus[duet][vocal: male, baritone][vocal: female tenor][ harmony]
call out to my love,
its amore
found we ar calling out
[Verse 2][vocal: female, tenor][response}
i hear you no babe i'm call you back
make sweet luv to me...
the let me cuddle you till the mornin light

suno does have choir and a few duet typesjust not really allow duets or harmonic choir settings: for vocals setting( ish have a bad asd day i'm drop words peace out lol)
so step to is to put something like this is description

genere: rock duet, r and b duet
mood: love , slow melody, and hamormy
productions
vocalsz: studio quality, clean, male lead baritone ,female vocals, tenor
Production: Clean mix, crisp diction, studio quality, dry audio,
male and females call and response, vocalists sing a harmony duet as the chorus lines

[quiet solar]

And my musings on the main practical routes to making duets, if it helps anyone:

Option 1 - I Feel Really Lucky (just using prompts)
Put that the song has two vocalists in the style prompt, and describe each in a way that you can refer to them in the lyrics box (e.g. Male bass-baritone, Male tenor). Consider adding call and response to the style prompt as well if they are going to alternate line by line.

In the lyrics box, tag the start of each line where the singer is supposed to switch over with something like [Male, bass-baritone] or [Male tenor]

Prepare for multiple attempts and using lots of credits - each changeover of singers makes success less likely.

Option 2 - Multiple Extends (ideally requires Pro or Premier)
This works well if the singers don't change over frequently.

As Option 1, until you get a song (or songs) with the two voices you want to use

Optionally: create personas using just the vocal stems for each singer

Fix any vocal switchovers not happening correctly by using Extend just before the failed switch, and prompting for the singer due to perform the next line using the Persona to help change the voice over and keep the voices consistent. If necessary, this may need to be done for every single vocal change. Continue until done, then use Get Full Song and remaster it to deal with inconsistencies in instrumentation and style between different Extends.

Option 3 - Two Versions, One Splice (requires Premier or using an external audio editor)
*This works well if the singers don't need to harmonise. *

Create the song as normal using one singer.

Once you're happy with it, either a) use Extend near the end of the song and re-create it entirely, or b) use Cover with very high audio influence to try to avoid key/tempo changes, this time prompting for the second singer that you want.

Extract the instrumental and vocal stems from both songs, and replace the vocal parts of singer 1 with singer 2 in Studio / your preferred editor as needed.

Option 4 - Master Version + Fixes (requires Premier or using an external audio editor)
This is your best option if the singers do need to harmonise.

Create the song as in Option 1 with both singers.

Keep going until you're happy you've got something reasonably close and, critically, that you like the instrumentals of. This will be your master version which locks in the instruments. Get the vocal/instrument stems from this song.

Use Extend to create 'fixes' for each error in the song. Disregard the instruments and focus only on whether the Extension fixes what you want in the singers.

Extract the vocal stems from each fix and use them to replace the vocal stems in the Master version in Studio / your preferred editor as needed.

[hard beacon]

wow.. i can believe my eyes.. no message in the prompt channel in the last 14 hours ?? 😄

[polar spade]

2-day hangovers?

[hard beacon]

maybe 😄

[hard beacon]

btw... there was a very very good video on youtube about duet (prompting and style) in the last 2 days.. maybe i can find it again..

https://youtu.be/26831gFFYQg?si=kHlEGkvNN2jsjrlI&t=56

Maybe this one will help some people with the duet thing 🙂

[molten jay]

I text with your structures, didn't work! always i have problem when my music has 2 vocalists

tested*

[hard beacon]

well.. for me it works.. also in the past... for me it wasn't new.. BUT.. i think it depends a lot of the style

[molten jay]

i wrote "[Verse 1 - Female Vocalist Only]" but male sings

[hard beacon]

for me it worked with "handmade music" like ballads or stuff like that.. EDM also works good for me... but Rap ... well... not that good

[charred trout]

My payment has been deducted successfully, but the subscription is still not activated. Please either activate my subscription or initiate a refund at the earliest; payment proof is attached for referenc

[molten jay]

mine is Dark Pop, Industrial Electronic

[hard beacon]

@charred trout : check https://help.suno.com/en/articles/2551617

@molten jay : in the end of the day it's still AI.. it's not working 100% 😉

[final hazel]

Please do not post any "proof of payment" because everybody can see your personal information. You need to write to [e-mail] and wait for their response. If it takes too long you can go to #archived-ticket-help and ask for help there,.

[quiet solar]

Honestly, Woodhouse. xD

@molten jay Assuming you described both singers in the style prompt, then your lyric tagging is fine and you've done all you can using prompting. The outputs will inherently be a bit random - if you create a few more generations, you'll find the singers will typically switch over in different places; not necessarily where you've said, or to the singer that you've requested.

This isn't a problem that can be fixed by tweaking prompts. If you scroll up a short way to my most recent post earlier about the "four options", that gives you three other practical routes you can use to get a duet as an end result, starting from the point you're currently at.

PS. For anyone checking Robin Clark's video - the prompting used in it is absolutely spot on, but the video maker is a bit misleading to suggest that it leads to duets that appear to be as controllable as in the very short example he gave, which just so happened to go absolutely fine. I suspect he didnt include any of the versions where Suno completely ignored him and did it's own thing. 😉

[earnest agate]

Has anyone been using EQ, music theory, chords, mastering techniques, etc., in their prompts?

Basic examples:

OR

Boomy low end

Tight bass

Forward mids

Shimmering highs

Muted top end

OR

200 Hz boost

3 kHz presence peak

12 kHz shelf

Mix of both?

Use the Styles field to persuade Suno in the direction you want it to go, bmp, music style, case chord etc..

Create a full song, or create a basic beat/outline/structure of a song. Take the generation you like, pull it into Suno Studio, and build the song mix using musically and technically phrased instructions (EQ, music theory, chords, mastering techniques, etc.) for each prompt in the “layer.”

[earnest agate]

LOlz That doc isn’t here to explain how it works, it’s just a bunch of techniques to play with.
Did anyone have a go? Any success?

[hard beacon]

like i said before... good results for me 🙂

[frigid orbit]

old man weakness

[frigid orbit]

numbers dont mean much, unless it's commonly metioned, like an era or popular bpm.

[thorny laurel]

Hello, I'm looking for a prompt for a singing style but I don't know much about music style, I think it harmonies I'm not sure
Basically like the first verse of this song https://suno.com/s/AuiD53zMtLXHHK67 with variations in the voice
I'm looking for a medieval fantasy singing voice

[earnest agate]

yeah, that was is the direction i am experimenting with now. had mixed results with more technical detail. sometimes amazing and others suno seems to loose it, no cohesion to the instructions

[quiet solar]

@earnest agate I haven't tested enough to be confident saying anything one way or the other, but I'm somewhat sceptical about Suno's reliability in responding to prompts when it comes to referencing highly technical details or very specific musical phrasing (e.g. "use this precise chord sequence") when it's referencing it's training materials.

It does seem comfortable taking a steer on elements related to mixing and mastering, and a general steer around musical aspects (e.g. major/minor, key signature, bpm), and that can make a huge difference on the perceived quality and direction of the output. But I suspect there's a level at which it will struggle to reference specifics (e.g. can Suno identify a 12 kHz shelf in the songs it's been trained on? I genuinely don't know).

Some people absolutely swear by highly complicated, technically detailed prompting - and if you find it works, then absolutely go for it. But in a number of examples people have been kind enough to share with me, I haven't seen Suno following those detailed instructions any more reliably than other prompts - and I've seen people use far simpler approaches to prompts that generate equally great outputs.

There's an inherent randomness behind all of this that makes it difficult to pin down, and I'm sure at times we're all guilty of thinking "Oh yeah, that latest tweak I made to the prompt REALLY worked" rather than recognise it might have been purely down to random factors beyond our control happening to come together.

And to really get to grips with it, we need to know how many 'failed' attempts there were behind that great sounding song to know if one prompting route is more effective than another overall. Our own individual experiences are bound to vary too much to be a reliable indicator alone. 😉

[quiet solar]

I'd be tempted to try including "epic" in your prompt, similar to that reference track. Suno described similarly emotive vocals in one of my songs as "operatic and powerful", so that might be worth a try as well.

So something like "Epic medival fantasy, performed by an operatic and powerful male vocalist" might be a useful starting point.

[grave heron]

I am pretty new to Suno and have been pasting lyrics in. Today I wanted to see if Suno could write lyrics but I don't understand how to make a prompt were it says "write some lyrics or a prompt" it just turns it into lyrics even if I say "write a song about" or something.

[fathom hemlock]

Normally just a short sentence and than click on generate text if you are not on individual you to click on the plus in the app Modus so that the text is coming up

After the 6 generation I got finally my rain. Suno is sometimes like a lottery 🙂

https://suno.com/s/vY4FMcoOl34bp8B4

[grave heron]

Ah ok I didn't see that button there to just generate lyrics.
I previously had it write lyrics when I didn't want them so I was confused about the interface.

[fathom hemlock]

Now you got it - easy

[charred quartz]

Anyone getting jitter, static, clicks and random interference in every gen? It is getting ridiculous now as I mustve spunked about 2000 credits now and every one is crap. I even tried trimming prompts down to a couple of basic lines and cleared browser cache but still awful. Only started hapening about 2 weeks ago. Any idea folks?

[fathom hemlock]

Do you have an example to share

[remote ivy]

whats your genre, and have you tried the following tags [style: high-fidelity studio mix, warm analog mastering]
[vocal_style: clean, detailed]

[grand pebble]

We have this same issue recently, but will try using those tags. Thanks @remote ivy

ok i grabbed this from my trash, but around 50-55 secs it is quite clear the "jitter": https://suno.com/s/M0rUlqVv8rjK0f7v

[remote ivy]

you might some examples in the workflow thread, or if you have a worflow you would like to share, that is a good place to put it

[remote ivy]

you would call how life was sung jitter is that what your talking about?

[grand pebble]

there is some sort of muffled click, or interference - dunno how to describe it really but we get it on every gen

here is another example @remote ivy line 2 of both 1st and 2nd verse https://suno.com/s/OE2ewU22xT699Zr6

[remote ivy]

aside from actually resolving the issue...have you had success w/ remastering or tried it if you had particularly good render w/ 1 or 2 glitches...i've done it about 5 times to salvage a render, but mostly it was cause weirdly sung word, or an actual glitch...a pause in a vocal run when it should have continued, discontinuity in instrumental , etc

[grand pebble]

yeah don't sweat is as this isn't tech support 😆 But yeah, used loads of credits doing remasters and covers and recovers but still get the same glitching. I raised a ticket but, well, its the season...

its such a shame as this issue only reared its ugly head about 10 days ago

gonna ht the sack now anyway. Even virtual rockstars need their sleep! Thanks for the suggestions @remote ivy

[fathom hemlock]

Yeah burning credits for nothing we are used at Suno 🙂

[rocky haven]

When writing song lyrics, if you are looking for an earthy, heavy, ancient feel, borrow one from Tolkien's book and use Germanic derived English and avoid Latin-Origin English. if you use Ai to help with your lyrics you can indicate this. to show what I mean, compare: "The illumination descends upon the mountain." (Latinate/French influence), With: "The light falls on the fell." (Old English/Germanic influence). Germanic words tend to be shorter, stress-heavy, and consonant-dense, with hard stops and clustered sounds that anchor the voice in the chest and give speech a percussive, grounded feel. Latin/French-derived words are often longer, vowel-rich, and smoother, encouraging legato flow and forward resonance that feels lighter and more ornamental. In the case of Tolkien, his “earthy” effect comes from stacking stressed monosyllables and compound nouns, so meaning lands through weight and rhythm rather than melodic smoothness.

[thorny laurel]

OK I'll try, thanks!

[reef pendant]

If we're noticing the same thing (I describe it as a "glitch" in the audio), this is something we're aware of. For now, although it won't perform the same (but should avoid the issue), try using "Style Persona (legacy)" by clicking on the cog wheel after selecting the persona.

[frigid orbit]

There's a cohesion if you just adapt to suno than hoping suno will adapt to you..

[pure panther]

Trying to tease out maximum whwomp whwomp

[high hatch]

Hi All, I’ve noticed that when creating songs with vocals, Suno often uses very similar male or female voices. Is there a way—such as a specific prompt or setting—to make the singer’s voice more distinctive or varied in how it sounds?

Also, for EDM-related creations, is there a way to prompt Suno so the track starts with bass before the other musical elements come in?

[pure panther]

This sometimes works to get it to start with bass (working on one rn haha)

[high hatch]

I am fairly decent with my prompts but it doesn't always do what I want and was wondering if there was a better way to prompt to acheive the specific reuslts I am looking for.

[pure panther]

I get some very very interesting vocals styles using descriptive terms like these. I'm often surprised how varied the vocals can sound using descriptors like this
female vocals Deep powerfully expressive female alto blues, full of scoops, slides, grace notes, conversational/sarcastic delivery and wide, lingering tight vibrato, exaggerated expressive inflections,

Here's a really fun one for male vocals: exaggerated theatrical delivery (tenor/baritone), wide expressive inflections, sneers, occasional scoops; “circus” tension with rock

[high hatch]

Also, when I want to be descriptive in terms of the vocal singer I am looking . Do I add that to the lyric or style prompt? I noticed beofe that if I tried to get creative and put some insctructions in the lyric promps it wouldn't do what I say but rather sing the lyrics or say them. Which is not what I want.

[pure panther]

I put it in the style prompt

[quiet solar]

You can do a mix of both. The style prompt is best for general steering for the tone of the song as a whole, lyric prompts if you want to influence a very specific line or section.

Generally speaking, style prompts are more reliable than lyric prompts. However, Suno shouldn't sing your lyric prompts if you put them in [square brackets] although sometimes rare glitches do happen.

[high hatch]

Noted and Thank you. can you provide a specific example just so I can be clear?

[quiet solar]

Sure. The most typical approaches people use are along the lines of either;

[Verse - soft instruments, passionate vocals]
your lyrics go here as usual

or

[Verse]
your lyrics go here
[passionate vocals] the lyric line you want to influence
more of your lyrics here

[pure panther]

@flint haven Why tho?

Woodhouse dislikes prompt tips in the prompt tip channel lol

Calculation:
At 138 BPM in 4/4:

1 bar ≈ 1.739 seconds

100 bars ≈ 173.9 seconds ≈ 2:54

---

[Intro – 8 bars, keep sub pulse continuous]

[Verse 1 – 8 bars]

[Verse 2 – 8 bars]

[Bridge – 8 bars, no break in bass]

[Refrain – 12 bars, hook]

[Verse 3 – 12 bars]

[Verse 4 – 12 bars]

[Bridge – 12 bars, no break in bass]

[Refrain – 12 bars, hook]

[Refrain tag / final line repeat — 4 bars]

[Outro – 4 bars]

In Style box:
4/4 ~138 BPM

Ok, it's less pretty but not banned lol. Using this has gotten me the closest to duration control out of anything I've tried so far.

[high hatch]

Hi All, when creating a song. Suno often comes back with songs that are short. Mostly under 3 minutes in length. How can I ensure it does at least 3:30 or more? Is there a specific prompt I need to use? of course I will tailor it to the Genre length but some guidance is appreciated.

[formal rain]

for me, it always depends on how i structure the song itself. are you doing like a traditional
Verse
Chorus
Verse
Chorus
Verse

?

[high hatch]

Yes pretty much

[formal rain]

try adding an Intro and an Outro. how much time are you trying to make up?

[high hatch]

At least a minutes worth

[formal rain]

sometimes ill use [instrumental break] and it'll fill for about 20 seconds

add those in between see if that gets you some more time

[high hatch]

Will do. Thx

[quiet solar]

Suno will generally take as long as is needed (up to the max generation time) to go through everything in the lyrics box. Once it gets to the end, it'll typically bring the song to a close.

So the essentially answer is "put more in the lyrics box". In addition to QH87's suggestions, adding extra verses and choruses is a fairly straightforward route. [Instrumental break] or the longer version [Instrumental interlude] or things like ['Your instrument here' solo] can be helpful for padding too, and you can put them pretty much anywhere between sections and Suno will generally go along with it.

You could also try the slightly more nuclear options of [Instrumental verse] or, indeed, an instrumental of any section that has played previously.

[lone narwhal]

Hey guys!
Usually I am reading only and learning a lot here which I really appreciate.
Now is the time I need to ask something (out of utter desperation ^^):

How to make Suno create a piece without drums or any kind of percussion?

My goal is to create an ambient electronic piece ...yes - without any kind of percussive instruments /drum machines.
Tried different ways of prompting
a composition without drums, percussion, or rhythmic beats. Minimalist electronic. only synthesizer pads, atmospheric sounds, and melodic elements. rhythmic feeling through harmonic movement and subtle sound wave interactions. ambient drone soundscape. Bright, clear and positive.

or

ambient, drone, breathing, pulsating, layered, glitch fx, deep, bright, no drums```

and another one I cannot find again. 
10 tracks all with drums.
How to avoid it?

[proud dagger]

Is there a page which contains names of all the instruments that Suno recognizes or have?

[sharp kettle]

drums drive the core of 90% of music, you can try in the negative prompt
also use the production: prompts: and [ tags ] no words but most music is going to have percussion and drums..

[lone narwhal]

yes - you are right and I'm fully aware of the drums. Tried negative ones as you can see. In within the first 30 seconds drums came in...
will the production prompt make a difference?

[sharp kettle]

Genre: electronic dance( post any other genre style references
Mood: fill in the blanks for me
Key instruments: ( fill in the blans for me)
Vocals: Instrumental (No Vocals)
Production: Clean mix, crisp sounds, studio quality, dry audio,
(fill in the blanks for me)
Tempo is 140 bpm.
Description: A pristine, studio-quality track, (fill in the blanks for me)

fil in the blanks
as for the lyric brackets i will write thos after you fill this in for me. to give you a basline to try. if it works it works>
in the exlcude zone you would have to type ever instrument you want excluded in brackets.

[lone narwhal]

Tried that before. Same result. Unfortunately I deleted it immediately out of frustration so I lost the prompts as well (bc not saved).
Therefore - do you know a specific way (prompt) to avoid drums/percussion?

I guess I have to go extremely specific, right? like describing every single beat, note etc?!

[sharp kettle]

fill that prompt in for me im going yo use it to write lyric prroduct structure prompt.

the at the bottoms i will give a you a baseline of exclude instruments, sinc e i don't know if suno has all \the instruments in your prompts

[sharp kettle]

i need more info to help you structure for attempts and give you a baseline idea lol unlike my most complex song my edm prompts usaly work in 4 attempts

[lone narwhal]

Mood: relaxed, dreamy
Key instruments: Synthesizer 
Vocals: Instrumental
Production: Clean mix, crisp sounds, studio quality,
Tempo: 90-100bpm.
Description: An ambient, minimalist electronic composition crafted solely with synthesizers, beginning in silence with a soft synth hum that gradually introduces a gentle, ascending melody. Synth layers build sequentially, creating rhythm through texture rather than drums. A distinctive "breathing" synth mimics human respiration, generating a relaxed, meditative atmosphere. The piece transforms electronic sounds into a living, organic soundscape, evoking tranquility and optimism through carefully layered synthetic textures that feel paradoxically alive and responsive.```
like that?

Or do I have to go a similar way like this example:

[sharp kettle]

error 1: you have production notes in the description, move those to production
example: creates rhythm through sounds rather than drums.
error 2. you are missing key instruments: breath sounds has a instrument name
error 3: your description should expand on the mood as a prompt..

give me a second but i can write you a lyric production lay out. but keep in mind drums ar 90% of music so ther is no qaurentee it wont have them, i also have to fetch the names of the instruments you named in the desrciption. and oh that an even better way for me to start lol

[lone narwhal]

Oh! thank you for pointing out my errors.

[sharp kettle]

303 bass is a type of drum line, so you have 1 drum listed already

[sharp kettle]

i see three drums its going to read

[lone narwhal]

just an example of prompting from another artist.

Holy moly... "breath sound has an instrument name"... 😄 what name is it?

Or you mean modulation? So it is "breathing" ...because that is what I intend to do, right?

... like sidechain-modulated pad or so? but is it in instrument then?

[sharp kettle]

[Instrumental Intro]
*[ genre ( recommend year to pull goes here)ambient electronic, ambient drone]
*[ tempo100 bpm][key Instruments: Synthesizer, talk box ][breathing sounds: br, HH]
*[Vocals: Instrumental][Mood: relaxed, dreamy]
*[production: An ambient, minimalist electronic composition crafted solely with synthesizers, beginning in silence with a soft synth hum that gradually introduces a gentle, ascending melody]
** [production line 1][ insert here]
[production line 2][Synth layers build sequentially, creating rhythm through talk box rather than drums.]
[Instrumental][Verse 1]
[place the * or ** here]
[Instrumental interlude]
[Instrumental][Verse 2]
[Instrumental][Bridge]
[Instrumental][Chorus]
[Instrumental][verse3]

make sur to space it and line it correctly. uo ccann also play with the bracket lines using production and decription line to guid it into a mood structure

ad for the negatives>
exlude styles [exlcude: snare drum}, [exlcude: 808 base drum], [exclude: 808 base drum][exclude: percussion]

[sharp kettle]

that above was for you

[sharp kettle]

talk box and yes that lay out i gave should help no promoises though lol

[lone narwhal]

Thank you very much for your effort! May I ask some follow-up questions?

[sharp kettle]

if yah need correction just fill out the templete and ill correct it would house migh censor though

[lone narwhal]

Where do I put it in? Sorry - I'm confused. Does it belong in the lyrics??! Or is it the same way for styles? (must sound silly ...)

[sharp kettle]

the break down for the lyric box is like this btw

[Instrumental][Verse 1]
[place the * or ** here]

[Instrumental interlude]

[sharp kettle]

[Instrumental][Verse 2]

[Instrumental][Bridge]

[Instrumental][Chorus]

[sharp kettle]

[Instrumental][verse3]

[instrumental outro]

sory forgot that braket

[lone narwhal]

so the 'lyrics' is used to give it a direction and order but the styles is for specification of above direction?

[still coral]

NYA NYA VOICE: glitchy vocals, autotuned vocals, processed vocals, anime vocals, kawaii vocals, light vocals, cute vocals, high-pitched vocals, pitch-shifted vocals, heavily processed vocals, vocoded vocals, synthetic vocals, digital vocals, metallic vocals, formant-shifted vocals, anime-style vocals, artificial vocals, neko vocals, nya-nya, vocaloid-esque, chipmunked, ring-modulated

[sharp kettle]

this runs the machine in prompts
Genre: ambient electronic, ambient drone
Mood: relaxed, dreamy
Key instruments: Synthesizer
Vocals: Instrumental
Production: Clean mix, crisp sounds, studio quality,
Tempo: 90-100bpm.
Description: An ambient, minimalist electronic composition crafted solely with synthesizers, beginning in silence with a soft synth hum that gradually introduces a gentle, ascending melody. Synth layers build sequentially, creating rhythm through texture rather than drums. A distinctive "breathing" synth mimics human respiration, generating a relaxed, meditative atmosphere. The piece transforms electronic sounds into a living, organic soundscape, evoking tranquility and optimism through carefully layered synthetic textures that feel paradoxically alive and responsive.

in the lyric box this gives ai machine directions
[Instrumental Intro]
*[ genre ( recommend year to pull goes here)ambient electronic, ambient drone]
*[ tempo100 bpm][key Instruments: Synthesizer, talk box ][breathing sounds: br, HH]
*[Vocals: Instrumental][Mood: relaxed, dreamy]
[production: An ambient, minimalist electronic composition crafted solely with synthesizers, beginning in silence with a soft synth hum that gradually introduces a gentle, ascending melody]
** [production line 1][ insert here]
[production line 2][Synth layers build sequentially, creating rhythm through talk box rather than drums.]
[Instrumental][Verse 1]
[place the * or ** here]
[Instrumental interlude]
[Instrumental][Verse 2]
[Instrumental][Bridge]
[Instrumental][Chorus]
[Instrumental][verse3]
[Instrumental outro]

its a double protection sitch for instrumenals and songs so you dont product alot of wasted creds

[lone narwhal]

Alright - I just follow your instructions to fix my errors (you mentioned above - I am grateful!) and it should work with Suno in a good mood

(got no creds left for today - try it when reloaded)

[sharp kettle]

this is a sample its my stream song of when suno gets it right. https://suno.com/song/d06a7476-a397-45f6-80df-a9bb0fa78a09.
as you can hear the rythm actuaal fits how i sing.

[sharp kettle]

feel free to tag ask me again in discord. when its done an tryed you may have to repeat my notes i tend to forget extra complex music notes lol. note pad for the win

[lone narwhal]

Thank you! I'm gonna do so!

[earnest agate]

cheers, i think i have come up with a hybrid solution which i am testing atm. Those 1000 characters in the styles field, are gold

[quiet solar]

Just to add to DrSupermanShadow's advice, for future reference - try to avoid negatively prompting Suno.

If you put "No Drums" in the style prompt, Suno will almost always ignore the no and read that as "Give me drums, please!". Similarly, it won't understand other conditional words like 'without' or 'only' reliably. You're best to limiting what you put in the style prompt to only the things you do want included.

[lone narwhal]

Thank you!

[earnest agate]

This is great. I think you’ve just saved me a day (or more) of work. I was trying to get the timing right using the [SEQUENCE:] tag, but it wasn’t great. The song timing was all over the place and I felt like I had zero control. I was just about to start switching things up. This is much better.

I got my little friend to knock up a document for me. Have a look and tell me if he’s talking absolute bollocks or not. Just to be clear, this isn’t the way to do it, and I’m not even saying it works yet. It’s a first draft and completely untested.

If anyone can sanity-check this or confirm it works, shout up. My little friend needs constant validation and can be a real pain in the butt sometimes. He even titled the document (Expert Level) 😄 Thanks for sharing that.

[sharp kettle]

suno is the king of butts when it come to time control. if found that its best to be pro and to hum a intro and outro. lever the interludes blank and sing out the song yourself... hit cover and have the vocalist replace you... and your beats out.
its usally spot on by 10-20 seconds but when it breaks it makes the song ether extra short or long,

if you have e drum pad its also best to tap the song beat out, although if yah can fight out audacitys beat function im sure you can tap record your figures and replace that with it, the sing the song, and cover it and replace the beats sections with real instruments lol

and my asd brain just when fart so sumbuddy can retype that in clear speech lol

[quiet solar]

@earnest agate Jury's out, but I'm leaning towards "it's not reliable". My sense when testing it is that it mostly respected the song structure, but not the bar count.

8 tests with bar information at 125bpm:
Duration range: 2:02 - 2:30. Average length: 2:19.

4 tests with identical structure and no bar information at 125bpm:
Duration range: 1:54 - 2:51. Average length: 2:23

It's a limited set of tests, and we know Suno doesn't take into account every single lyric prompt. The bar structure may be having an effect (there was less variation overall in tests with it - note: very small sample of data), but it's not precise enough to sit down in advance, calculate how long you want the final song to be and built the bar count accordingly.

I think you might need to have a chat with your little friend. That kind of overconfident "you've told me this is true, so now I'm going to make up a plausible reason for why it might actually be true" is a bit of an issue! 😂

Labelling it as an expert guide and not saying anywhere "This is all based on guesswork by me, an AI!" risks being a frustration trap for newer users who won't understand why they can't get the results they expect despite following the confident "expert guide". Then they come in here and complain Suno is broken... 😁

[fading cargo]

Hi all. I would like to ask how can I make the lyrics to be like “whispered”. I have looked up many videos where they say you need to use brackets and such to make it work but nothing seems to work for me. Any ideas? Also, how can I make the intro of the song instrumental and after a short time the lyrics to start? It seems to ignore any instructions regarding Intro or instrumental build up and such. Thank you in advance if you guys can help!🙏🏻

[quiet solar]

If you're trying to have a single line whispered, it's a bit random. You've probably already tried variations on putting [Spoken: whisper] or [whispered vocals] in the lyrics box before the line you're trying to affect, but that's the right sort-of approach for prompting it.

I've not really tried to use whispers enough to say if there's a particular phrasing that works more reliably than others. All I can suggest is turn the Style Influence up under Advanced Settings (if you're a pro/premier user) and keep trying.

If you're trying to have the whole song whispered, then you could put something like "soft, whispered, breathy vocals" in the style box.

[fading cargo]

Thanks for the quick reply. I have tried that as well. Spent over 800credits last night only to try to make it work😅 what I’m looking for can be whispered or like far away in the background like a soft humming sound. I got only a few line of lyrics that have to be “whispered”. It seems to ignore any of my instructions even on high style influence. Since I cannot import copyrighted music as inspiration, how could I do it?

[quiet solar]

Do you have access to Studio, or are you comfortable with using a DAW / audio editor like Audacity?

If yes, you could record yourself whispering the lines, put that audio into the relevant section of the song, then use Cover - it should replace your voice with the AI singer, as Suno is pretty good at taking audio cues for things that are tricky to prompt.

[fading cargo]

That a pretty good idea. I will give-it a shot later today. I do not have the full plan, only the Pro. What about the instrumental intro part? How can I delay the lyrics to start?

[quiet solar]

The general approach would be to include [Intro] or [Instrumental Intro] in the lyrics box - if Suno isn't playing ball, sometimes it helps to try adding more of a description in the lyrics tag and/or the style box, e.g.

Lyrics
[Intro - soft instrumental, synthesiser and bass]

Style
The song starts with soft synthesisers and bass, gradually building up to the verse

[earnest agate]

Yeah, I guess I probably shouldn’t have left the (Expert Level) in there, but I thought the light-hearted way I posted it, joking about it was enough to convey the intent of the document. Just trying to help give people options and ideas for getting to their vision of what they want. As we know Suno has no template and never will have. That is what keeps the coins falling through the slot 😃

You’re never going to create a template for prmpting and that’s a good thing. That’s where the "creativity" comes from, and why each of us doing slightly different things can (and does) get different results. All I’m trying to do is give people ideas and options to help them get closer to the idea or vision they have, within the many limitations Suno has. Or rather, the limitations we have.

I’m under no illusion that we’re the test dummies or guinea pigs. We put money into the slot machine, that money helps keep the Suno elves paid and the lights on. We’re not going to get access to the model(s) that will be given to large corporate entities to create the next hit singles and superstars, the billion-dollar music catalogues that will recreate the Beatles, or something like them, in some form.

This is just a bit of fun. For me, that fun is figuring out how to squeeze as much as I can out of Suno and creating the kind of wacky songs I like. I’d love a 10-minute, ’60s/’70s-style psych/prog rock fusion jam sessions like they used to do back in the day. Fifteen minutes of pure craziness that somehow produced incredible sounds.

So that doc was created from a bunch of things I’ve tried, or had lined up to try, and all I really did was ask my Suno friend how this (Tea_theLeast suggestion) might work to solve the [SEQUENCE:] tag issues I was seeing. And again, he didn’t disappoint. There were a few things it came up with that I hadn’t even thought of trying.

Incidentally, talking about Suno interpreting things: I think Suno works out song timing using a few main factors—BPM, singing style, instruments. It gets the song length from those, regardless of you trying to specify timing directly. And it has to, because most people just ask for a “1980s rock song” or “’90s punk,” and Suno chooses BPM, singing style, instruments, number of verses, intro, outro, etc., and creates a song length it thinks is appropriate. That said, it is also capable of letting us nudge it in different directions using various techniques.

[quiet solar]

Oh, it was really clear from your post that it was an experiment, and you didn't know if it worked.

My criticism was of your AI friend, in that it should really have taken that part into account in what it fed back to you. 😉

[thick hornet]

You know, I personally miss the presence of tags in Suno (like in html) with opening and closing ones, like <baritone voice> text, text, text </baritone voice>

[earnest agate]

Yeah, that’s the thing: unless you’re corporate and spending big money, you can’t guarantee a workflow or income stream using any of these LLMs or AIs. They will always change on you, and you’re back to square one making sure the workflow still works as it should and that nothing broke.

It’s a system designed to keep you running on the wheel, making stupid amounts of money for a small group of people. A first in history, at this scale.

[polar spade]

@oak maple

Just responding to your ticket-help question.

PayPal as well as Apple Pay, Google Play and others are pay processors, so yes, they charge extra ( a surcharge) for their services. They basically are supposed to provide a bit of security to validate you and give Suno confidence to authorize the purchase. So they charge an additional fee on top of the regular subscription price and pass it on to the user, you. There also may be currency conversion fee for doing that as well.

[oak maple]

Thanks for the info! Even though I understand the reason now, I don't understand why the surcharge is around 2€. That's an upcharge of nearly 20%. This is huge!

[polar spade]

Yea. Its a big hit. I agree. Kind of paying extra for convenience when we dont have a credit card or workable debit card. There may be info on charges in PayPal agreement of service and service fees.

[oak maple]

Hmm, I don't know if convenience is a valid argument in this case. As far as I know, PayPal requires a linked and functioning credit or debit card in order to pay with it, and you also have to confirm your identity in order to use the service. And yes, that makes paying online and now everywhere else much less complicated. Nevertheless, I have never seen or had to pay such a surcharge anywhere in order to pay with PayPal. However, if this surcharge is really due to additional fees from third-party providers, then it would at least be fair to customers to make this clear during or before the payment process, wouldn't it?

[polar spade]

All good points.

[spark citrus]

Hi Guys, joinning in to get some pointers.
i'm 5 hours deep in a song and can't get my head around it.
music is okay tier but i can't get the little push needed to make it great.

do some of you have a bit of free time for a quick check ?
I must have messed up some prompts at some point

[quiet solar]

That sounds a little vague, but if there's something specific you're trying to do then I'm happy to help if I can.

[spark citrus]

Thanks, i'm making music over anime and then build an amv on top of it.
I'm working on escaflowne and man this is way out of my confort zone.

Goal is the escaflowne itself with simple song path: awakening - march - battle - victorious - rest

So far my best results are :

music is good already but barely A tier it miss a bit to get the exitement and battle type for the amv building on top of it

[quiet solar]

Hmm...that's more of a musical style question than a prompting one. If what you're not happy with is the sound and feel of the track, it's a little tricky to know how to advise you, since I don't know what's in your head, and you've clearly been making structural and lyrical changes as well as stylistic ones.

What I can say is that your prompting structure (especially lyric prompts) is absolutely fine, but very detailed. Possibly too much so, as that can mean Suno is less likely to pick up or focus on any specific steer that you're giving. Similarly, you're excluding a lot of things that probably don't need excluding. Most of the steer for the song as a whole will come from what you use in the style prompt, which Suno tends to pay a lot more attention to in my experience.

I'd also avoid including prompt instructions in () brackets, as if you're using those to also indicate a backing singer or a difference in how Suno should sing that line, you stand a reasonable risk of Suno singing your prompts.

The narrative flow you're looking for in the music is an interesting challenge. If you're not happy with how it's representing certain sections and you're struggling with giving each one a detailed steer, one option would be to break the song up, e.g:

Write prompts focusing just on the first section until you're happy with it, then Extend from it and change up your prompts to be focused on what you want from the second section, etc. When you get to the end, use Get Full Song and then cover or remaster it as needed to smooth over any rough edges.

One other possibility is to try generating the underlying track in v4.5 rather than v5. The two can have quite different musical interpretations, and you can always cover a 4.5 melody in 5 if you prefer the vocal style that 5 brings.

[spark citrus]

thanks for the tips and the time you took for me, much appreciated.
I'll try the generation section per section. Great idea !

[oblique token]

I need some help. I tried adding a prompt with and without lyrics, but the problem with this prompt is that it always highlights the guitar or piano at the beginning. How can I remove this? I tried again with a different style, but it's still the same.

[quiet solar]

@spark citrus One final point - if you feel the song as a whole lacks the orchestral swell and power you're looking for, that can be more of a mastering/production issue. v5 is pretty solid with mimicking mastering and production effects, but unless your Suno profile is the final destination for the song, you might want to consider downloading it and tweaking that further yourself.

If that's even further beyond your comfort zone, Bandlab has a passable AI mastering function online that it wouldn't hurt to try out.

[spark citrus]

i'm with audacity 🙂

thanks 🙂

[quiet solar]

Good man. 👍

[spark citrus]

could you share the tracks ?

[quiet solar]

@oblique token What do you want to happen at the beginning of the song? An intro that specifically doesn't have a guitar or piano in it all, but is led by a different instrument?

[oblique token]

@quiet solar At the beginning of the song, I want the effect, but when I specify that I don't want guitar at the beginning, it still plays, and with persona too, because I'm testing them.

[quiet solar]

Okay. I'd echo Laguna's point - this might be easier if you share an example song here so we can see what you're currently doing with prompting.

It is generally a bit more difficult to tell Suno not to do something, and you're usually better off telling it only what you do want it to do.

[fathom hemlock]

Any idea how to restrict in Suno that on the beginning should not come up an own singing intro.
After 20 times of generating the shared link it has not done it.
Or better what is the trigger that Suno is not creating an own into singing? In the middle of the song it is sometimes Ok in the shared song it fits but also hie can this be prevented. Additionally faster the music as lighter is the voice dark voice is getting really complicated

https://suno.com/s/fyKyiGFpphfAHngE

[earnest agate]

Attempting a moody vinyl scratch with a jumpin’, skipping kind of vibe.
https://suno.com/song/81897e41-ed15-4cd1-ab3a-721bbf823468

[polar spade]

Building on what @quiet solar noted, I suspect the parenthetical remarks of the instrumentation are not perceived by Suno as an explanation of style towards the instrumentation, but a desire to augment the referenced instrumentation or anything coming before or after it by those style set in parentheses.

So a "[Deep guttural male choir drone only, distant taiko heartbeat pulse (soft percussion entry)]" does not just ask for the taiko drums to start out soft, but its asking for a soft percussive entry of everything within: the male choir, heartbeat drums , whatever is in brackets - ultimately making a guess at it or ignoring most and prioritizing anything its already previously doing.

Taking things segment by segment if simply to observe how each reacts when adding to the next I think seems paramount. Even moreso, less can be more. Honestly though, its hard to tell everytime, so pushing limits and experimenting is fine. But I would hold back on parentheses inside of parentheses or brackets. Just my observations.

Put in simple terms, its not good to mix all colours at once, or you just get a murky brown. Mix a couple, dab the canvas, see how it looks. Clean the brush, mix a couple more and paint again. Blend areas, scratch and build others. Eventually you have a colourful masterpiece.

[spark citrus]

thanks 😉
Following @quiet solar idea i removed the ( ) and reworded the command differently the difference was notable, then i combined cover and persona from previous succefull try and reworked final product on suno editor.
Result is very close to what i envisioned so this works for me 🙂
(now finetuning it on audacity ^^ )

[sleek stump]

Just a quick note, Suno generally ignores BPM commands, it will use Tempo rates common to the genre(s) you have detailed in the styles. It does however generally honor traditional tempo terms such as Allegro, Andante, and Presto, as well as fast, medium and slow.

As always, YMMV, Suno like to do what "it" wants most of the time, which is dependent on its training data.

[quiet solar]

v5 seems a bit better at respecting prompts for specific BPM and key signature in the style prompt - certainly a lot more so than older models. Though I suspect it's strongly influenced by genre and our constant friend 'randomness'.

[sleek stump]

I have had no luck using BPM commands in v5, but the traditional terms work for the most part.

[quiet solar]

Literally a perfect example of YMMV. 😂 Also agree that your suggestions tend to work well across genres and models.

[earnest agate]

Yeah, for me as well, BPM seems to be one thing Suno picks up on when it’s entered in the styles field. Those 1,000 characters we have to play with are the most important, I think anyway. Even still, it’s really just giving Suno a suggestion of what you want. I’ve pulled tracks into studio that were prompted at 90 or 100 BPM, and the two generations can end up with different BPM numbers. But the track speed is basically what I want, and I can always change it in the studio if I really want to keep the song.

It’s like trying to get the instruments you want. I tried to get a nice, clean harp track. what a pain in the butt. A harp solo, only a harp, no other instruments. Basic prompting to start: solo harp, Celtic harp, etc., etc. I had nothing in the lyrics field other than basic structure just to get an instrumental only output. I was still getting guitars, violins, and other weird stringed sounds.

I would hope Suno has training data for a harp, so it may just be a case of phrasing or using the correct harp name in the styles field. I can’t wait to try and get an authentic theremin sound.

[polar spade]

@wispy mulch One thing is you can do first is paste your lyrics to chargpt or other and ask if they recognize anything offensive. You may need to ask that this would include words formed from partial letters between words like "finish it" or words found inside of words , such as "harassed".

[dire zephyr]

Any prompt wizrdard that can help me straighten out a lyric GPT prompt to be used for suno. I lost my thumb drive will all my prompts and 3 years worth of work and Im almost back where i need to be. I need to finalize and make GPT understand how to humanize lyrics using the enlish language and to write with imagery mispellings cadence rhyme types and basically how to write like a mc using language arts the correct way for structure. I feel like GPT is stuck on stupid or got dumber.

[frozen dagger]

Mondegreens and slant rhymes... ask it for that

[molten jay]

google play is easy, fast

I need help guys, i want a kiss (not from you, from Suno :))
I used this [Kiss sound] but didn't work, I used this [Sample: (muah) - Quick kiss sound on every 4th beat], didn't work either!

[covert forum]

Can I get help, please, on how to prompt of a British villain?

I have a song I really Like, I added my own vocals (female) over it and uploaded it, and I want Suno to give me that song back, or close enough, with just a change in the vocalist. But I don't know how to prompt in a way that keeps the vocal melody the same? It keeps putting in high rises, or swinging the melody in strange ways, or doing the intro English and the singing American... I'm trying to do distinguished evil villain, but the best I'm getting is "20-something brat". Any help would be appreciated, please.

I keep trying baritone British male, English accent, menacing, villainous deep... I don't know so I'm here hoping for some direction.

[polar spade]

@west fjord On listening through twice and skimming as well, i wasnt able to hear click and nothing consistently occurring. The most i could hear was the last 10 seconds or so what would be similar to a drum kick not properly miked giving a plosive sound. So this might be a click sound in your playback that doesnt occur as strongly as in my playback.

Tried stemming out the tracks and see if any one stem has the clicks?

And I had zero stuttering.

Can you pinpoint a specific few seconds where its the worse?

[molten jay]

How to i tell Suno, i don't want any voice like hoho haha aah ahhh at first of music?

[frozen dagger]

Was working on my PC, then I switched to mobile phone. Songs that played fine on desktop were clipping and stuttering on mobile device. Noticed it earlier at work but dismissed it. Maybe same issue?

[wise kindle]

they're called adlibs, so you could try adding that into the exclude styles section

[molten jay]

I did, but again and again i have that in first of music

, **No reverb on hook, ** powerful pop-scream, **NO ad-libs, NO humming, NO "ha ha" sounds, **

[wise kindle]

using negative prompts is a bad idea. if you're on pro or premier I'd recommend using the exclude styles instead - it works a lot better

instead of "no reverb" you might use something like "dry mix" instead

by negative prompt I mean don't write "no <word/s>". It generally doesn't parse that correctly and will think you want it instead

[molten jay]

gotcha Thanks

[wise kindle]

no prob. just in case you weren't sure, this is where I meant to put them

[west fjord]

Here is a much clearer example of the issue with Model v5:
https://suno.com/s/NmleXBjztui2z8Zr

I have pinpointed the exact timestamps where the clicking/stuttering artifacts are most audible. It happens frequently throughout the track:

0:06, 0:10, 0:52
1:10, 1:15, 1:18, 1:40
2:05, 2:08, 2:13
3:20, 3:35, 3:51
4:01, 4:05, 4:17, 4:21

Please check these specific spots. It sounds like digital popping or interference that isn't present in previous models.

[crude jetty]

Anyone else having issues when you put " <something> Instrumental" even with lyrics, creating a instrumental track?

[wise kindle]

"instrumental" is a pretty strong, high-priority word. I wouldn't use it in a prompt unless I specifically wanted an instrumental

[crude jetty]

I ended up using "instrumentation" and it seemed to behave better

[wise kindle]

yeah that's a smart move. glad you got it worked out!

[crude jetty]

but the idea was to take the clean instruments and muddy them

[polar spade]

☝️ First, @west fjord are you listening on Phone or Desktop. Have you see this response? by OBLMOUS

[frozen dagger]

i also listened to the most recent track that was shared on my phone, pc with speakers and headphones. was less noticeable on pc speakers. makes sense. Could hear the clipping at the 6.5seconds and 10 sec and 1:10 pretty clearly. It reminds me of playing a guitar and my cable jack has a loose connection,

im hearing this in my own gens now. =\ just heard 3 static breaks (within 20 seconds)similar to the artifact error being shared. like a static rip, or "jack pop" like i was saying

[crude jetty]

share something perhaps? multiple ears on something would make sure it's not just a you thing

[polar spade]

@west fjord Yea, on desktop I used headphones, and heard most of them, adding 3:55, 4:27 and a bit muffed one at 4:37 Sound like a hioshigi stick with a sock over it. yea, like a cable pop. Yea, You can mention in the bug room then, see if its a temporary glitch but if others are getting it, then a bug they should look at. https://discord.com/channels/1069381916492562582/1082775596767117312

I cant stem it out for you, so certainly an interest to see if splitting the stems makes a different and if so if any one frequency/instrument.

If in the original then try a remaster just to see if it was a bad few hours of the day, and remaster may recognize it as an artifact and "clean it up" , as well.

[frozen dagger]

meh im having a couple of issues tonight. im thinking it is high traffic problem rn. weekend. lots of new users form holidays. tweaks on backend effect us without us knowing all the time. I would love info on Suno traffic, like peak times to avoid. See it it helps.

[crude jetty]

I don't hear popping, what I hear is a layering issue. also there's no negative prompt which could help the situation

[quiet solar]

Suno isn't specifically trained to produce sound effects, but it can occasionally surprise us.

Either keep trying (and best of luck), or look to add it in editing using an actual kiss sound effect - obviously, ensuring there are no copyright or intellectual property issues with doing so.

[molten jay]

Just a kisss! (Kidding) 🤗

[frozen dagger]

I'm not looking for tips to fix, just sharing to show that it is not an isolated issue for @west fjord. Clipped from gen I made earlier tonight. like 5 or 6 pops.
https://suno.com/s/iRoRxQNZInktSQhN

[crude jetty]

the thing I am hearing is just pitch changes, not popping

[wise kindle]

<@&1082792303976001547> one here too sorry

[final hazel]

No problem.. we need your help atm with these scammers.. Thank you so much

[frozen dagger]

got my headphoens on and the first one is at 00:04 its pretty clear, but yeah. remaster, recreate, split and clean...

[crude jetty]

I have a 20w soundbar

[steady crane]

Played with unique instruments in this. Results..Flexy AI 😅

https://suno.com/s/02Gm8MPqQinp3NjZ

[crude jetty]

mods gettin'em before I can even finish the reports lol

[steady crane]

Was trying out stuff like Oberheim OB-Xa

[crude jetty]

pretty cool

[steady crane]

Albeit my singer kinda muffles some of those sounds lol. Can definitely tell instructions there but would prob sound clearer in instrumental

[crude jetty]

here's a Theremin and Guzheng track I was playing with a bit
https://suno.com/s/jYAvxdinPKh90IlK

[quiet solar]

So in this exercise you have two separate challenges - 1) getting the accent you're looking for, and 2) keeping the melody unchanged. I would strongly recommend addressing those in two separate steps.

Starting with the vocalist - personally, I've found prompting for accents to be a little hit and miss as well. If you want to keep at it with prompting, try bass-baritone rather than baritone if you're looking for a deeper voice. Adding phrases like 'theatrical' and 'character performance' may help, to try to pull in examples from musicals.

Alternatively, I've converted two British accented vocalists I have into Personas for you that you're welcome to experiment with. If you favourite them, you can select them in your own Create screen under a sub-menu:

https://suno.com/persona/52151fba-6e62-42ce-aed2-2e496cc21a93
https://suno.com/persona/132c5762-71fe-4544-8c43-eaa469a34da4

Once you've got a vocalist you like from experimentation, create a persona for them if you don't have one already, and cover the original song using that Persona. If you have Pro/Premier, you'll want a high-but-not-too-high audio influence (enough to keep the melodic arrangement, but not so high it clings to the original vocalist). Try setting your sliders to something like 5% weirdness, 50% style and 70% audio - though you will probably need to tweak them to find the sweet spot.

Hopefully that's clear - let me know if not, and let us know how you get on.

[polar spade]

Suno is not liking the 1:00 to 2:00 spot tonight. Glitchy on 80% 4.5 and 5.0

[west fjord]

<@&1372996242325508176> @frozen dagger Thank you both for testing and validating this.

To answer your question: I have tested on BOTH Desktop (PC with headphones) and Mobile, and I hear the stuttering artifacts clearly on BOTH.

It is not just a mobile playback issue for me.
Also, this happens on every single track I generate with Model v5 (v4.5 works fine).

It seems like a deeper rendering issue with v5.

[crude jetty]

modified vocalist: https://suno.com/s/CcoBaTiBPEcPHk6u

original:
https://suno.com/song/481584a4-5e4a-4dfa-ba00-4eb39312f837

something like this?

(note the song choice might not have been the best as it's more of a cheerful tone than villainous)

[frozen dagger]

Remaster was enough to clear up most of the popping in the one gen I messing with still, but that's still not much comfort. Hope it smooths out soon.

[frank matrix]

I want to remix a Malay-language song into a tropical or upbeat tropical style, but why does Suno always turn my Malay-language songs into stupid TikTok dangdut? Please fix this problem.

[crude jetty]

it's likely a prompting issue, what is the full prompt you are using?

[frank matrix]

" catchy tropical house genre" that's the only prompt i used in style

[spice whale]

use the exclude prompt, describe what you dont want, like no ticktok stuff
Even better describe the voice that you dislike musically

[crude jetty]

yea, that's too vague

[spice whale]

Or more flat like:
No cringe, no rap sung text, ....

[frank matrix]

i tried put SLAP HOUSE in style but Suno still gave TikTok's remix

[spice whale]

Hmm give me one sec, i dont know the ticktock style but let me try to create an exclude

Could you describe what you mwan with ticktock style?
Im not on ticktock xD

[frank matrix]

It's the song that Indonesian DJ remix on Tiktok

[spice whale]

Helps me 0 xD

[frozen dagger]

if you have not tried already, I would recommend working on the music first, then add vocals. would ask myself if there is something that exists that is close to what i want and use that as a reference starting point. The Malay lyrics added after the music is made may help tho.

[frank matrix]

much like this type of beat https://www.youtube.com/watch?v=WmcKU8Iwnqw

[spice whale]

SUNO MUSIC STYLE (Slap House, but not TikTok-core)
Slap house / bass house, 124–128 BPM, tight sidechain pump, punchy kick + crisp clap, bouncy offbeat bass, modern EDM polish, short vocal chops as texture (not hook), dark-bright club energy, clean drop design, minimal lead melody, big low-end but controlled, wide stereo FX, festival/club mix

[spice whale]

SUNO EXCLUDE STYLE (anti–TikTok / anti-“snippet bait”)
exclude: tiktok trendbeat, “viral” pop, hyper-bright happy synths, cutesy bubble-pop, sped-up / nightcore, chipmunk vocals, whistle lead, ukulele / pluck-pop, generic motivational chords, overly catchy singalong hook, talky influencer vocals, meme SFX, handclap-only grooves, lo-fi bedroom pop, overused riser-slam formula, short 15s drop structure, “dance challenge” rhythm

[frank matrix]

let me try

[spice whale]

Something like that xD

[crude jetty]

Generic TikTok music is a short, hook‑driven, loopable audio bite that combines a catchy melodic fragment, a simple drum‑bass groove, and often a vocal or sound‑effect punchline. It leans on clean, bright production, operates at moderate tempos, and is designed for easy reuse across countless short‑form videos. Its popularity is sustained by the platform’s algorithm, a low barrier to entry for creators, and a self‑reinforcing cycle where trending sounds spawn remixes, challenges, and brand‑new “generic” tracks.

[spice whale]

Its not my music direction, so this is just a very wild guess withhin my own trained ai parameters

[spice whale]

Thats basically pop and most of music xD

[spice whale]

Parially xD

Let me know if it helped in any way^^

[crude jetty]

that's what Nemotron said so... maybe it's somewhat accurate?

[frank matrix]

okay 🙂

[spice whale]

Nemotron Whats that?

Wait i can ggoogle xD

[crude jetty]

A local AI model

[spice whale]

Hm dont know it, so i cant say anythign about that

[crude jetty]

It's fairly competent, and somewhat light on the hardware

[spice whale]

I dont know wich databases it uses and how it interprets music, atm i am my own tool with music leistungskurs in the abitur xD

[spice whale]

Hmmm can be, i realy dotn know man xD

I have my own musicall background, and am sus about stuff i never tested myself, so i take your comment with natural skepsis^^

[crude jetty]

you can searc this on goggle for info: Nemotron-3-Nano-30B-A3B

[spice whale]

Is that one realy one for music? XD
https://build.nvidia.com/nvidia/nemotron-3-nano-30b-a3b/modelcard

[crude jetty]

I didn't say it was a music model.

[spice whale]

And well my local hardware is not bad, but it cant compete with something like the bigger datamodells xD

[spice whale]

All fine just pure inreterst^^

[crude jetty]

but it can identify and descibe music

[spice whale]

Ai nerd here^^

[spice whale]

hmmm but music analysation already does that?

Have you doublechecked a music song like lets say tchaikovskys first piano movement, gave out your ai prompt and cross prompted that with a musical analysation?

Might do that xD

[crude jetty]

dabbler here

[spice whale]

I mean i might do that xD

[spice whale]

We all are xD

[crude jetty]

I mean, this is accurate

[spice whale]

I agree

But i would even more expand the explaination

of the elements listed so suno can interpret it better?

OKay endless iterator here xD

[crude jetty]

there's a lot more nemotron said, I didn't want to post a massive picture

[spice whale]

easy^^

Well create a song then and post it^^

[pure panther]

I've been cleaning up more random clicks and pops lately too. Most of them are audibly unnoticeable, but you can see them in Audacity.
I'm finding them mostly in the vocals

[crude jetty]

https://suno.com/s/8ksx9P1c5099IFfJ

[spice whale]

Listening...

Melodical for sure

[crude jetty]

it takes a bit to get into the grove

[spice whale]

Not bad, try a close remix for better sound quality

[spice whale]

I know what an intro is xD

[crude jetty]

I mean it takes more than just the first verse

[sleek stump]

frankly, if it takes longer than 30 seconds to get a listener interested, the likelihood of them hearing the whole track drops.

[crude jetty]

That's fair, but it was an experiment in prompting directly, not the length

[spice whale]

Do you make music for the masses tho?
For the guys on shorts that cannot listen to more then 2 mins or not?
If you have no intro you wont hook me^^

[sleek stump]

The answer would be yes. I've written songs that were recorded and charted by other popular artists.

[pure panther]

My shorts (less than 3 minutes) can get thousands of views (very small channel), videos (greater than 3 minutes) ...not so much. Even though, overall, the videos are better quality.

[spice whale]

Cool, i have nothing against that^^

[spice whale]

Hm but going for that trend enhances the trend, so i myself cannot participate

Its not what i want, more and longer complexity in song and text

[pure panther]

Haha, I feel you. I just put up what I think sounds the best, short or not. That's why I have a mix.

[sleek stump]

Its the short attention span world we now live in.

[spice whale]

For a while we will

[crude jetty]

@spice whale https://suno.com/s/JVRs5FgWFPPZlLZh better?

[spice whale]

But i predict that one falls too xD

[spice whale]

Listening...

[pure panther]

Uh, yeah, but when that happens, will anyone being listening to digital music?

[spice whale]

Thats why you get clicks and i dotn hahaahahah
My marketing is more then abysmal, not that i give a duck

[spice whale]

Thats already happening xD

[pure panther]

Same, I don't market. I let the bird fly free and whatever happens happens. I make it for me first anyways.

[spice whale]

I even forget to let the bird fly sometimes hahahahahah

[crude jetty]

I got one more, it's a female singer though

[spice whale]

To be honest, i like the female voip more^^

[pure panther]

The female voices generally sound less robotic imo

[crude jetty]

gonna check2 more gens first

[spice whale]

Hm that depends on the grenre and what you have put into the voice prompt part xD

[crude jetty]

https://suno.com/s/XDDE5fXhSkXdJQCp

@spice whale there ya go^

[polar spade]

Right. Thats good. Yea, this is the bump the turntable glitchy thing going on. Sounds like the needle jumping for a minute long. 🥴

[pure panther]

Oh, yeah, that's no good.

[crude jetty]

is it only from studio, or pure generations?

[polar spade]

Oh me? Its just first generations.

[crude jetty]

that's strange, because I'm not having that issue at all myself

[spice whale]

Hm tryed to remix it?
MAybe that filters it out?

[crude jetty]

have you tried to add "clean vocals, clean instrumentation" perhaps?

[polar spade]

Oh not at all. Just trying some prompts out. No worries, mate.

[polar spade]

Thanks for the suggestions. Not a serious endeavor on these songs. I can run another generation later. I always assume its a server issue.

[crude jetty]

The thing is nothing I have generated all night has had any issues what-so-ever

[spice whale]

Saaame
Hm maybe because of some gerne influence or crossover that makes it glichy?

[polar spade]

Hazzah! Great for you. No worries. Really. Just making a comment that I got some really harsh gens. Its all good.

[crude jetty]

well, maybe you could provide your style prompt and gender?

[pure panther]

I don't think I've been able to say that on any night 😂 Although, I set the weirdness pretty high and ask Suno to some very unusual things.

[crude jetty]

this high?

[spice whale]

Hahahaha

[polar spade]

OK. Router has restarted. Going back to my VR. L8tr y'all.

[spice whale]

The song only can go as weird as the promt allows tho
And then it still must be somewhat music
Okay we also have expressionism and such so "still beeing music" has a very high sealing

[pure panther]

I tried 100% once, it was the stuff of nightmares.

[neat nexus]

Dang I was hoping to get your thoughts on a redo of that song from earlier's ending. But have fun with your VR! 😄

[crude jetty]

lol I did that once. got a 30 second thing

[spice whale]

Aint the challange now to do 100% and dont get a nightmare?

[crude jetty]

it breaks completely if it goes over 85%

Been there, done that lol

[molten jay]

I tired, I burned 400 create to generate a song, everytime starts with he hehe or hahaha or ... why?

[remote ivy]

sounds haunted!

[lone narwhal]

hey - me again 😇
messed around a bit to find out how to create an ambient track without drums.
It seems I still have errors in my system?!
First img I put it in the lyrics window - 2nd img in the styles window.
Nothing works. It starts almost immediately with drums.

3rd img ... I tried to use prompting in the lyrics section ...

What am I doing wrong?

@molten jay I know it is silly and funny - please explain to me... what am I missing?

[molten jay]

I laughed, because I'm stuck like you, and since last night until now, it's 9:50am, I've been saying in every language, "Please don't play the lullaby music first, but it's still playing."

[spice whale]

Hmmmm

No intro lullably?

[molten jay]

for ambient i tried too before, one way i found is, i generated a piece, then i cut just that section that has no drum, then made it cover and again generated from that cover.

[spice whale]

Or just cut it out? xD

[molten jay]

intro

[earnest agate]

Did you try switching form v5 to 4.5, 4.5+ (if on Pro)? When trying to get something working that is just not, on matter what you do? i switch between them just to check if it's not a version issue before making a lot of changes.

[molten jay]

i did, honestly 4.5+ did not sing that haha lala 🙂 but V5 has excellent quality more better that V4.5

4.5+ listened to my all commands that was unbelievable

V5 doesn't but has better quality

[earnest agate]

lolz yeah thats the trade offs you working with atm and for even prompt is random getting v5 to start as you want. try stripping eveything back, get an into that works then build off that

[molten jay]

Now I have an idea. I'll cover that music generated with 4.5+ and generate it again with 5.

[earnest agate]

yeah, that could also work

[plush field]

NEVER use words like NO, WITHOUT...

you should put words you want to skip in Exclude style, and even there NEVER use negatives

so...

this is a proper way to remove things

if you do not have paid plan, you can try using - (minus) in style, but i am not sure how well it will work

so, something like this

...Clean Mix, Mastered, ‑Halftime, ‑Breakdown, ‑Silence, ‑Trap Beat, ‑Pop Ballad,...

similar things apply for lyrics, do not use this

[Chorus - Fast & Chanted - No Pop]

use something like this instead

[Chorus - Fast & Chanted - High Pitched Vocals]

this is just an exaple, use what you actually need

another example
[Verse 1 - Male Rap - Fast Flow]
[Chorus - Female Vocal - Driving Kick]

insted of

[Verse 1 - Male Rap - No Pop]
[Chorus - Female Vocal - No Slowdown]

[lone narwhal]

Thank you! Have a free account only 😊
It is super confusing... one group is saying you need to use negative prompts and the other group telling me the opposite.
I've listened to creations with negative prompts which worked out as intended...other not at all.
I am aware that using any kind of genre will trigger a "preset" - so I avoided it...

For today I burned my creds in the "learning" process

[plush field]

the issue with prompts like NO POP is that model will most likely see POP instead of NO POP

it might work from time to time, but it should be avoided

model is not human, tokenizing is not same as human reasoning

[lone narwhal]

tried that as well - just using instruments or vocals...but Suno is putting in drums instantly.

[plush field]

if you want give me an example and i will create modifed version for you and send you the link right now

[grand tree]

This may sound crazy but I am crazy anyway – How do I get Suno to reproduce a very low quality reproduction – My aim is to reproduce a quality of a 1930’s 78 RPM playing on a gramophone I told Suno to BPM: 62
STYLE: 1930s dusty 78 rpm ghost jazz, muted trumpet, upright bass, brushed snare, detuned piano, soft trombone, ghost clarinet, heavy hiss/crackle, needle jumps, wow/flutter.

[lone narwhal]

as "simple" as: create an ambient track without drums. only synths layering and creating a pulsating, breathing rhythm. no percussion at all

[plush field]

do you have an exaple i can modify? might be faster

[grand tree]

https://suno.com/s/v6CGCTDcsABGh98Y

[lone narwhal]

sure - this is my 1st attempt...pretty naive prompting so a lot of room to improve
https://suno.com/s/lhMVO7YP7CWJPMWc

then I kinda fafo'd 😄
https://suno.com/s/9NVcb69IKL5rcVIO

[plush field]

so you are trying to create ambient electronic with 1930 sound quality?

[lone narwhal]

nah 😄 that's @grand tree with the 1930's song idea

I want to create a drum-and-percussion-free-ambient-track

[plush field]

ooops, ok

1 by 1

[plush field]

https://suno.com/s/9tqmAolurWqIrSkx

sounds almost right, except it miss cracking

[lone narwhal]

@grand tree that would be 4 u 🥳

[plush field]

https://suno.com/s/cWbfZuxJZFkcgzYr

here it is 🙂

[lone narwhal]

thank you! Now I see the "cheatcode" 😉 I need to use "-" to avoid drums and stuff instead of "no" or similar

[plush field]

let me try this with 4.5-all

[lone narwhal]

you probably doing it with the sliders, right?

[plush field]

i put those in exclude

[lone narwhal]

yep - cannot do it with the free account so may be that's the "workaround" for me then (to use "-" instead of negative prompting)

That was helpful @plush field ! Appreciated!

in addition I am going to use [section] instead of [verse] for the next attempt

[plush field]

nope

🙁

https://suno.com/s/GYQLpSKk06qNZOTM

[lone narwhal]

crazy!

[plush field]

with - in style and 4.5-all model 🙁

but, i also increased the style influence in the v5 prompt

with free, you can use only 50, 50

[lone narwhal]

interesting though! I need to find a different approach then I guess... I strongly believe there is one solution (for free accounts)

[plush field]

https://suno.com/s/ne5fPQ1i7dKoT4yP

almost

[lone narwhal]

"beatless" ... that's a good one! It adds some brushed hats in the middle though but nice!

[plush field]

btw. since you are on free account, use my UUID for title trick to supress getting very similar track each time

#📚┃suno-school message

[earnest agate]

anyone know if [effects:] | [fx] | [sfx:] are the same thing or are they supposed to do diffent things?

[plush field]

based on https://github.com/stayen/suno-reference/blob/main/tag-reference-llm.md they should be used for different things, but i think its not too important, and that both would do the same thing if you use them in a same way

sorry, that was for @earnest agate

of course, i meant if you use them for SFX effects like mechanical hum

but heavy reverb should not be used with sfx

[earnest agate]

Yeah, I see that there is a distinction, which I think now makes sense. At least it looks like i have some direction in what i was playing with. Cheers for that. A

[plush field]

but... there is always some but... 🙃 you need to use something in style which will trigger the effect

for example, this will probably do nothing

[Verse 1 - Deep Female AI Monotone]
[Voice Effect: Cold & Robotic]

unless you add something like this in style

Robotic Monotone

most obvious example is something like
[Turntable Scratch]

wich will not be triggered 99% of time unless add something like DJ Battle, Scratching or Vinyl Scratches in style

for something like this

[Intro - Dark Storm Atmosphere]
[Sound of heavy rain]
[Loud Thunder Crack]
[Low Rolling Thunder Rumble]

you need something like Thunderstorm, Nature Sounds or Rain or whatever

so, you can get those without using SFX as you can see

sorry for a typo, english is not my first language (obviously)

and my language accepts multiple negations, even 4 in the same sentence 🙂

[earnest agate]

Right, this is starting to make a lot more sense and i should have looked at styles, because the more one uses Suno, the more you realize that the styles field is the most important. Thanks for clarifying.

[earnest agate]

🙂 looked good to me

[plush field]

my english i mostly good, but sometimes i catch myself using some things that are not allowed in english, like using multiple negations, excluding the articles... 🙈 that happens if i do not think in english, but in my native language and then translate 🤣 always thinking in the language you are speaking/writing at that moment is the only safe way to avoid those stupid mistakes

[spice whale]

I dont know was you are talking über

Is inspiration via songs from playlists always random?
Atm i am kicking the older ones out so the right ones get choosen
Is there a way to manually pick em?

[plush field]

inspo = random from playist
persona = exact song influence

[hollow hare]

Inspo = Random 4 iirc

Not sure if that changed

[spice whale]

Limiting to 4 in a style playlist = not random anymore
Perona... 30 secs only... i find it limiting

xD i normally have 2 playlists, one with the songs and one to only throw in and kick out the songs i want to influence

[plush field]

there is nothing wrong with having just 1 song in playlist approach

[spice whale]

xD 4
Okay obviusly i start an albung with only 1 song, but when i reach song 6 i kick out song 1 for influences
Song 2 when creating song 7
And so ooon

Jeah well ill write suno^^

They can programmin a selector

[quiet solar]

Official steer from the Suno devs is to stick to around 3-5 songs in the Inspo playlist for best effect

[polar spade]

I made a playlist of the better ones [ out of 70 outputs] generated to pick prompts* that might work. @plush field had additional prompts which may present a different vibe as well: Foley, Field Recording, Found Sound, 1930s Jazz, Dark Cabaret, Phonograph, Mono, Degraded Audio, Wow and Flutter, Vintage Recording, Lo-Fi, Narrow Bandwidth

I only used a few excludes with mine so , maybe the more excludes, the better nostalgic it may sound > (these are from @plush field entries) ‑High Fidelity, ‑Stereo, ‑Clean, ‑Crisp, ‑Digital, ‑Modern Production, ‑Polished, ‑4k Audio, ‑Reverb, ‑Wide Stereo oneshttps://suno.com/playlist/c575ab86-9047-4b05-a260-02eb172016a6

Of course these days, its maybe better to work on the vibe and vocals and just stem out the foley sounds, crackle, pop, hiss and lay down as a separate track to accentuate the whole piece while EQing the vocal and instrumental to be more muted and mono sounding. So, either Studio or using your own DAW.

• oddly at one point, i spelled "gramaphone", and yet, those came out the better ones. 😁

[plush field]

using studio or external daw will definitely give much better results with less effort when we talk about effects, like cracks, statics...

but it is amazing how v5 reacts to some prompts and gives very close vibe to what it was asked to give

[polar spade]

Yea. Just doing best to keep things old school for those who just have 4.5all or Pro. Id use Studio these days. 😁 3.5 was good, but it does seem 5.0 accentuates the nostalgia better. But "lo-fi" and vinyl is picked up by 5.0 as more hip hop, than 3.5 did. So its about running 50 outputs and rotate the prompts until something clicks.

[plush field]

also, we should keep in mind that v5 is still in beta, so it is possible that some things will be even better in final version

[plush field]

https://suno.com/s/WFxczRb7GVaenpuf

looks like you got nice static and cracks in the beggening of this one, even i am currently listening on laptop speakers. i will check later with proper equipment (studio monitors)

[polar spade]

Im not the one trying to do this, yea? 😁 Yea, I'd make the foley stuff first and just senquence it all together, these days probably.

In then end, whatever mine sound like, the exact prompts wont work for @grand tree u. So its all about experimenting.

[plush field]

after i heard this i am getting an idea to make something like synthwave but with 1930 sound quality, might be very interesting 🙈

30s synthwave from alternate universe 🙃

[sleek stump]

Are you referring to the order in which your tracks are listed in a a playlist? If so, you can easily change that. Edit the albums playlist, and just drag and drop the tracks into the order you want.

[frozen dagger]

u can drag and drop individual songs directly from your workspace to the +audio +persona + Inspo area and then drop them in the Inspo to choose ANY 4 songs, wwithout the need to place them in a playlist.

[polar spade]

I didnt make the playlist public but left the songs remixable , right? Do you see the remix buttons on them in case you want to take some of the sound effects? Ill delete the post soon. @grand tree as well.

[quiet dune]

Is there away to do Remix's of songs from like music artist ?

[hard beacon]

for example?

[quiet dune]

@hard beaconi guess like david guetta

[hard beacon]

ok.. well.. Since you are not allowed to upload copyrighted material, the question answers itself 🙂

In other words: No 😉

[sleek stump]

Ya thats a giant no lol

[hard beacon]

Due to the Content ID system, I also assume that the upload will fail 🙂

[sleek stump]

If they allowed that, they would be fighting lawsuits night and day and need to hire 300 more attorneys

[hard beacon]

Well... at least it would create new jobs 🙂

[sleek stump]

It would also increase the cost of our subs!

[hard beacon]

For that, there's the “Ultra subscription” at $50 per month.

😄

[sleek stump]

😬

[frozen dagger]

Just want to add my 2 cents. The Suno scenery is set to change. Signed deal with WB in November, and then further than that now it looks like Netflix is purchasing WB. So over the next few years, I would expect some changes to the way things are handled. Possibly dramatic changes, even if rolled out over time.

[glacial tree]

Is there a bracket for this? How to get in suno a speed up high pitch voice as a sample? Anyone can send me brackets?

[pure panther]

Warner Music Group has partnered with Suno, Warner Bro's is being sold to Netflix. Warner Bro's is a subsidiary of WMG, so its sale to Netflix will have slim to no effect on the partnership deal with WMG.

[frozen dagger]

could try something like [Transpose vocals +5 semitones] or [High-pitch Vocals], but best method would be grab a vocal you like, pull it in Studio (if you can) and then just use transpose to do it yourself and make it a persona

[frozen dagger]

Just saying. Seeing speculations about different things. Not all directly affecting Suno use. Butr the media landscape in general. For Suhno specific I saw something mentioned about limiting our downloads. =\

[pure panther]

They clarified that it will be limited on the lower tiers, but not on premium.

[frozen dagger]

well that's good, i guess. but i will wait to see what happens.

[sleek stump]

For clarification, Warner Music Group (WMG) is who Suno signed a deal with. They are NOT owned by Warner Brothers/Discovery/WBD. The deal they have with WMG will not be affected by any Netflix deal with WB.

WMG has not been part of WB since 2004, Time Warner sold it to a private investment group and has been independent since.

The Suno scene is about to change, yes, but we have no solid details on what the limits are going to be for subscribers and on what plans they will be limited to.

[frozen dagger]

Just saying. Changes may be coming. Media corporations are absorbing each other like never before. EA, Microsoft, Bethesda, Disney, Hulu, Netlfix, Suno, WMG (and the weird TikTok "ban" that affected more than just TikTok (lost acces to Marvel Snap when that happened for a lil bit). Stuff like that. Can't reliably predict the future, but borderline monopolies and large corporations are the new norm. Which is counter-intuitive (imo) to the idea of user-generated content. But certainly some ripples on the way from these splashes.

[neat nexus]

If we're going to talk about the TikTok "ban" let's not forget the way that impacted the video editor CapCut by having it be down for a while.

But eh...this is probably the wrong channel for that. Try #🗣️┃chill-chat

[frozen dagger]

yeah, i dunno what all was affected (effected¿ never get that right) MArvel Snap was me. Lots of money sspent in that app and then poof was gone. lol

[hard beacon]

lyric prompt insight of the day: Suno has a problem using the word “CALM” after the word “DJ” 🙂

[frank violet]

Curious whether any music‑production grads are using Suno as part of a professional workflow. If that’s you, I’d love to hear your stuff and chat.

[sleek stump]

Thats all just business as usual for these giant media corporations. Disney owns almost everyone now, Microsoft has acquired a few big studios, EA has always gobbled up studios only to close them within months. This is all just normal for decades.

One more detail here, as of late 2025, Netflix is no longer interested in acquiring the streaming side of WBD, citing anti-trust red flags. For all intent and purpose, Netflix is out of the picture.

[hard beacon]

ok... now i am frustrated... 3000 credits later and the missspelling is still in the vocals... maybe i have to write them new to create a better cover version 😉

[neat nexus]

Sometimes that's all you can do. If replaces can't fix it, go back up the chain and extend from before the error and make the change.

[hard beacon]

the problem is, that is it in the pre-chorus that is 3 times in the track 🙂

[sleek stump]

I have found it light years easier to just run a cover and correct all of the lyrics there.

[hard beacon]

but i think it's a problem because of the original one is from v4.5 and also there is the missspelling...

[sleek stump]

Both 4.5+ and especially v5 are good at picking up corected lyrics in covers.

[hard beacon]

it wasn't in + 🙂

[sleek stump]

What I have been doing with my lyrics to prevet that from happening is to look at every word that can be pronounced differently and just phonetic spell those words from the start.

[hard beacon]

AND i try to cover into another genre... So there are various variables that need to be changed.

It's driving me crazy... the vocals are great, but the rest is rubbish... well... let's just leave this track alone for a while 🙂

[sleek stump]

There was a song I was trying to produce where Suno just refused to cooperate with me on.. It got so frustrating that I pretty much killed the project. Some things just arent meant to be 😛

I think that was in 4.5, maybe I'll go see if its still in the trash and see if I can resurrect it in v5...

[hard beacon]

forget it.. waste of credits 😄

[sleek stump]

lol, it certainly was back in the day... 😄

[agile plinth]

They said that there will be zero downloads for non paying , and limited downloads for every one else.

For people who think Suno will survive the WMG take over, WMG took over Udio, when this happened Udio sounded better than Suno (this was with Suno v4

Now Udio sounds worse than Suno v3

Ai models get better with more training data.

Reducing the training data will kill Suno.

If you read the lawsuit, it becomes clear that they are gonna kill the algorithm.

[pure panther]

The post in the announcment channel used to say that Studio would get unlimited downloads iirc, it says it's been edited so perhaps it's being walked back? Just to be sure I wasn't loopy, I found another post affirming that.

[sleek stump]

Its really not worth worrying over whats coming, we dont have any REAL details on the changes, only snippets of whats "planned". Wait until actual details are released by SUNO (not WMG), then you have clear info to make your decisions with. Speculation is never good.

Remember its a business deal with WMG, they are not taking over.

[pure panther]

Speculation is useless. I'm trying to stick to the official narrative. If it all crumbles, I'll be no worse off than I was a few months ago, and $30 dollars a month better off than I am now 😂

At least from a financial perspective.

[fresh gull]

Since I'm new here, I'm asking has anybody else figured out how to do? Multiple voices properly. Like dictating. And they got really weird on me. So I just let it random now. On whether or not it goes male or female. On a verse, even when I. specify

[dawn tusk]

@fresh gull it’s difficult isn’t it? In a remix contest it worked fine for me. But it’s pretty random

[fresh gull]

I had it working real well then I woke up and they broke a code in it - last month I think they inverted or added a - symbol someplace - went from perfect to complete wth!

[fresh gull]

I was getting really good with it when writing epic rap battles then it got wonky end of november

I have figured out how to get kids voices and such no problem and it great!

[dawn tusk]

Maybe you can help me then 🙂

[fresh gull]

@dawn tusk here is my style block ok:
Style: Phonetic Trap Bass House, Playground Chaos Dubstep, 178 BPM, massive punishing sub-bass 28-52Hz, nursery rhyme corruption aggressive, children's game energy meets festival chaos, playground chant wobbles distorted, innocent surface crude reality, sing-song delivery aggressive bass drops, schoolyard energy chaotic delivery, nostalgic corruption bass assault
Vocals: Call-and-response playground chant energy aggressive delivery, gang vocals schoolyard crowd chaos, whispered conspiracy sections knowing delivery, backup vocals multi-generational nostalgia corrupt, innocent children's game aggressive adult awareness, deadpan delivery phonetic trap execution
Mood: Playground nostalgia corrupted chaos, innocent game crude reality revelation, childhood memory aggressive bass energy, sing-song surface chaotic delivery, nostalgic corruption euphoria

[dawn tusk]

I have two question I’m also new here. 🙂

I specifically state [spoken word] but it is abandoned mid text. Curiously it happens with female vocals. Male vocals are more consistent for spoken word.

Also, no matter what I try I can’t get the female voice to be on the lower register (contralto) it always goes to this annoying breathy girl voice as a default

[fresh gull]

Spoken Word is hard it wants to SING - i would try open poet mic style

[cyan jewel]

been struggling with this too. can't get the j-core/denpa vocals i'd like if I'm trying to do sharpnelstyle stuff.

[dawn tusk]

I specified G♯2 to F5 as a range nothing. Always the breathless whispering girl

[fresh gull]

Ok I think I know what you mean @dawn tusk check out this (I got deep gospal voices here)
Style: 1950s bus engine idling as beat, civil rights gospel choir vs. authoritarian monotone, seat creaks as percussion, quiet defiance vs. rigid rule
BPM: 70
Mood: Quiet rebellion, "no" as revolution, tired feet vs. uniformed authority, passenger vs. driver, one seat that moved millions

I did Rosa Parks vs The Bus Driver (SOOOO GOOOD)

[dawn tusk]

Worked well for me with male voices

[fresh gull]

Also Ethnic Designations help

I wont spam the style blocks Check these out for help on voices (I would explain more but dogs need out atm)
https://suno.com/playlist/8edfcc78-22c8-4132-b0ba-57eeeb84134d

[dawn tusk]

I’ll try that. It’s strange how this biases are so much gender based

[cyan jewel]

getting anything but pop vocals out of female voices is difficult in v5.

[dawn tusk]

Should I try previous models?

[cyan jewel]

i usually do 4.5+ and then move it into v5

v4.5+ is WAY better at more unique vocal textures

esp. for female vocals.

try doing female harsh textured vocals like growls, screams, etc in v5, then try the same prompt in v4.5+ and notice how much better 4.5+ is at just going ham

[dawn tusk]

Thank you for tip!

[quiet solar]

In current models it's possible to prompt for multiple vocalists, but not to reliably control how they're used in the song and when Suno switches between them. You can give Suno stage directions for it in the lyrics box, and that helps a lot, but you can't instruct it to follow your prompts precisely - the more times you switch singers, the more chances there is overall of a failed switch.

What you're experiencing isn't the code suddenly breaking, it's always been this way. It seems luck was smiling on you until November. 😉

[fresh gull]

I wish you would remove 7:59 Gen Caps LOL I can write 12 min songs

[quiet solar]

You can extend from near the end of the song and continue it on. The 8 minute cap only applies to individual creations, not to the final product.

[cyan jewel]

extend it. thats the point of extend. i've done a few longbois

[fresh gull]

It use to not cap I did 11:13 single gen

[cyan jewel]

the gens are supposed to cap at 8m but sometimes it outputs a longer track. very rare and hard to intentionally cause to happen

and is nearly always worse than manual extension

[fresh gull]

TY i never thought of that I would see where it got cutoff and be like oh well - now I what just add back what it skipped?

[cyan jewel]

since the ai has issues with attention span and quality deeper into the gens

particualrly past 2:30 and 5

[quiet solar]

The length of the cap can be influenced by various factors - if you're giving it Audio feeds, using Personas or have a style prompt that stretches to the end of time, that tends to result in the cap kicking in earlier.

[fresh gull]

Depends on your "Directing skills" I think but could be wrong

[cyan jewel]

i find a good low volume point to extend for cleanliness, and then delete the portions of the prompt that got skills.

[cyan jewel]

its both. the ai model is known to have issues with quality post 2:30. directing can only take the model's own inborn limitations so far though

4.5+ was the worst you'll see a whole hecking block of rectangle for loudness

like oops all compression

[fresh gull]

Ill experminent then, I have had very good through put with 9 min songs

but I try to keep 5 to 6 mins

[cyan jewel]

mine turn into giant loudness capped heavily compressed rectangles lol

[quiet solar]

@fresh gull Coming back to your question about duets/rap battles - if you have a clear vision for the song and need to specify which singer sings which part, there are ways to do it, but not through prompting alone. You either have to get involved in a little bit of audio editing, or tactical use of multiple extensions to build a final version.

Mini-guide of the options available to you here, if that helps: #📚┃suno-school message

[dawn tusk]

Studio prompts helped me in this case

[fresh gull]

I havent gotten into the break and fix methods yet I will tho - soon

[fresh gull]

I have been doing slight alterations and reruns for now LOL

Ok this is serious question [ ( " - I know those make a difference right?

like to echo , back, etc...

[dawn tusk]

I find them to be inconsistent I.e [spoken word]

[fresh gull]

Example:

[SECTION 1 - WA ASSAULT]
(180 BPM relentless phonetic attack)
(Bass grinding catastrophic)

Wa-wa-wa-wa-wa-WOwoWOWOwow! (WOW!)
Wa-wa-wa-wa-wa-WOwoWOWOwow! (WOW!)
Wa-wa-wa-wa-wa-WOwoWOWOwow! (WOW!)
Wa-wa-wa-wa-wa-WOwoWOWOwow! (BOOM!)

[frozen dagger]

(parenthesis) best used for when you want lyrics sang as backing vocals. [brackets] for non-lyrical instructions to the ai for promtping in the lyrics box.

[fresh gull]

ok so instruction there trips up got it

[frozen dagger]

sometimes parenthesis will work as if they were brackets, but the way i said is the more reliable syntax

[fresh gull]

This helps me out because I have tried many cominations and still see things is ignores or says when not required

I learned never to put " "'s in style block unless you want everything after it read out

[frozen dagger]

sometimes using "quotation marks" around a lyric within your song will trigger a different vocalist. especially if you have a line break (space between lines). worth noting as well. can act as a call and response kind of thing sometimes.

[fresh gull]

TYVM for insights

[fresh gull]

I have done that and experimenting with things like (echo: word word word) or (chorus: "Lyric lyrics")

BUT IF YOU WANT A LAUGH MAKE A SONG WITH THIS!!!!!: (Adjust to music but its the PLAYGROUND and CHILDHOOD) trigger voices like tiny tim! lol
Style: Phonetic Trap Bass House, Playground Chaos Dubstep, 178 BPM, massive punishing sub-bass 28-52Hz, nursery rhyme corruption aggressive, children's game energy meets festival chaos, playground chant wobbles distorted, innocent surface crude reality, sing-song delivery aggressive bass drops, schoolyard energy chaotic delivery, nostalgic corruption bass assault
Vocals: Call-and-response playground chant energy aggressive delivery, gang vocals schoolyard crowd chaos, whispered conspiracy sections knowing delivery, backup vocals multi-generational nostalgia corrupt, innocent children's game aggressive adult awareness, deadpan delivery phonetic trap execution
Mood: Playground nostalgia corrupted chaos, innocent game crude reality revelation, childhood memory aggressive bass energy, sing-song surface chaotic delivery, nostalgic corruption euphoria

[frozen dagger]

that is probably a more reliable way to do that. In my song working on now line,
"And all that's left is the echo of the muse.
(muse, muse, muse)"
works very well

[fresh gull]

OK mental break ZZZ time

ttyl ✌️

[quiet solar]

Prompts definitely help, but at least for now can't give you 100% control in practice over a complex song with lots of changes between vocalists.

In those situations, if you really need to be precise about exactly who sings what, it quickly becomes easier to use a mix of prompting and editing to get to the end result, rather than trying to get it working with prompts and luck alone, even though that's extra work in itself.

[dawn tusk]

Yes, editing and exporting to a DAW helped me. Though honestly right now I’m not even asking for much just a contralto voice that speaks 😀