📚┃suno-school

[patent leaf]

I mean i have no issue going into a DAW and manually doing noise removal and autonmating EQ n shit to balance

[valid dagger]

Same problem here...

[hollow hare]

Jonathan Fly mentioned (#💬┃general-chat message) a current work around is to extend the stem and end it right away, downloading the combined music in the end

(I usually use "end, short" as part of the style tag on the extend-termination to try to get < 8 sec for the credit refund)

Alternatively, you can wait a day or two for the team to investigate and fix the issue 👍

[fluid sentinel]

What's the "extend-termination"?

[hollow hare]

A fancy way of saying it is an [end] extension, which people typically do to end/clip off songs in v3.5:

Extend where you want to end with the new extension set to either v2.0 or v3.0 when using ending tags (e.g. [end] in lyric box and "end, short" as the style tag) as they are much more reliable for ending than 3.5.

(Short to help end it quickly in hope for credit refund)

Of course, if you do not mind spending the 10 credits, you can just extend it in 3.5 and hope it ends. Regardless, it will need post-processing

[fair adder]

is there a way to reuse a music we liked but we changed the lyrics a bit in the meantime? becasue when i create it naturally generate an entire new song, i don't know if you can somehow cheat by using extention and then putting the time stamp at the begining

[hollow hare]

Aside from Audio Upload, not at the moment that I am aware of. There has been mentions of in-filling/in-painting for 4.0 but that is likely going to restricted to songs generated with 4.0+ due to the changes likely needed for its implementation.

You could keep trying to extend but that is more of a gamble than audio upload. If you were unaware, the AI studies the last one (for 3.0) or two? (For 3.5) minutes of the track when it tries to continue off.

IMO, best way to go about it would be to trim the section you love and want and use Audio Upload for that particular bit.

Maybe extend it as an instrumental piece so you have an easy start off point? I would use the instrumental tags (or even just [instrumental] or the lyrics) rather than the switch, especially if there are vocals present that may cause hallucinations

[untold briar]

If you mean "change the lyrics only a little bit", for example, change a few words only. Then this is what the future "inpainting" or "infill" feature should be able to do.

As Paimon suggests, sometimes you can use carefully chosen "Extend From Time" to accomplish this. Or, if you can edit the audio yourself, you can edit a piece of the song, "Upload" it. By choosing the correct segment of time in the Upload you may be able to recreate the same verse with the modified lyrics. (But then you still need to put the song together yourself at the end, using an audio editor. The "Get Whole Song" feature will only use the Uploaded section and be missing the rest.)

If you mean "generate a whole new song with new lyrics, but similar music style". While this is not a Suno feature, as Paimon also suggests, sometimes you can get lucky and make this work by "Extending" from a clip that has a sample of the music and probably a short voice sample. And a lucky Extension might produce very similar music, with new lyrics.

@fair adder

sometimes i wish that v3.5 was as creative and surprising as v3 cause right now all i can get to finish my song is something almost flat and repetitive
or pretty nicely turned but with a messed up chorus even if it is clear in the part i'm extending from

Let me check if I understand the problem. When you finish a song in V3.5 two different things can happen:

1. The end of the V3.5 song is boring and flat.
2. The end of the song is not boring, but the chorus in this ending is wrong. Meaning, the chorus is not sung the same way as the previous chorus in the song.

[fair adder]

no no, the first happens more often with v3.5 but the second happens with v3. v3 is more versatile but seems to have issues following the previous prompt, but now i think i can understand why since it only analyses one minute instead of all the song as i was originaly thinking

[untold briar]

@fair adder (I am replying to a message from a couple days go, I should have mentioned)
Continued:

If that is the problem there is not an easy solution. But some information that may help if you understand.

When you Extend a song you can either Extend a "Clip" (part 2, part 3, part 4) or you can Extend the "Full Song" that "Get Whole Song" menu creates. If you Extend a Clip, then Suno only has the musical memory that is in that specific clip. So if you extend Clip 3, if there is no chorus in Clip 3, then when you prompt the Chorus to repeat again in Clip 4, the chorus will not match. But if you "Get Whole Song" on Clip 3, then Suno will have the memory of the whole song as far back as possible (2:00 maximum) so even if the Chorus is in Clip 1 it will match. As long as it is within last 2:00.

V3 is the same, but instead of 2:00, it is 1:00.

[heavy geyser]

Hi, the stem separation feature has come available. But I get a time-out error every time I try to download the stems.

More importantly, the vocal track is just an extraction of the whole song, instead of the clean vocal track on its own. This means that the quality is mediocre, has artifacts and is not very usable to use in a DAW. The quality is just too poor due to the extraction method.

I assume that the vocals are a separate track when a track is built/mixed/generated, so is there any chance that the totally clean and really separate vocal-track will become available in the future?

[untold briar]

Okay. I left my reponse in chat, because it still could help by explaining in extra detail. Even though you already know. 😄

[fair adder]

yeah th'at what i figured out. i feel a bit stupid now cause it's in fact pretty logical xD

[untold briar]

Oh, I do have two ideas about V3.5 being boring.

1. The style prompts in V3.5 are overall "weaker" when compared to V3. So you have to repeat words, add dashes between compound words, repeat extra things like [big finish] in lyrics also. So maybe V3.5 is boring because prompts need to be different.

2. Because V3.5 has more musical memory, it is more consistent. It remembers exactly how to sing the chorus. But, sometimes being consistent means V3.5 is boring. It's like V3.5 is afraid to change the song, so it just keeps things the same. In these cases, it may be better to use less musical memory on purpose. Typically the way you do this is by Extending from a Clip, and choosing a time that is closer to the start of the song. This can make the music less consistent, but if you want the music change, then that is not always a bad thing. Then after the music changed you can still "Get Whole Song" on that latest Clip. This puts all the music back together. And you can choose an Extend From time in that new "Full Song" that is after the music changed for maybe 5 to 30 seconds. So when you Extend from there, Suno changes the music, but doesn't change it too much and it still remembers the previous song up to last 2:00.

Number 2 is long and probably confusing to read, but really is not complicated to do. Basically just generate a small section of music by removing Suno memory, so it is willing to be more creative. But then only use a small section of the new creative section so the overall song still kind of makes sense, all together.

[fair adder]

oh now i generate, extend then stitch, then i do it again, even it i end up only keepint like less than one minute of next generation lol

[untold briar]

This is a good default way to work. But for times when you struggle to change the music, then consider the opposite. (Extending from a clip and also Extend From times that are closer to the start of that clip)

The extreme example of this is if you want to make a song that is crazy changing styles. Like it is 20 seconds of country music, then opera, then k-pop, then rap. That is almost impossible to do, unless you use very short clips to kind of make Suno 'forget' about the previous style of music.

[fair adder]

i also wish so badly that we could like add a timestamp of song section we like so the ai can replicate it next time! i had to get rid of a very pretty extention because there was a whole minute that was unusable for me, and the part i liked was way farther in the clip

[fair adder]

i'm personaly not the one to do that kind of experiments but why not xD

[untold briar]

I ask Suno for feature to do this and more all the time. Fingers crossed (I can be pretty annoying)

[fair adder]

xD

the ai does a nice job already but we want to participate too! xD

[untold briar]

I love those experiments the most, myself. It's truly unique music!

BTW I know many users with screen had problems with the new Suno design. If you notice problems you can let Suno know in bugs-browser channel. You can mention me and I'll make sure that feedback isn't missed at least.

[fair adder]

well, as long as i can navigate with my screen reader i'm fine xD

[untold briar]

Well... the day after a new update sometimes mistakes happen and just that might be impossible for a little while (this may have happened before...) 😄

[untold briar]

For prompting music with a change, using a Stem to Extend it can make some pretty interesting variations that still smoothly blend with what came before. Soesn't work for suno website song links, since you have to manually put the pieces together of course.

[cold zenith]

thank you for the stemp

but i here some noise in the stems

hear

[untold briar]

Instrumental Stem One Time VS 10 Times

(If you Extend a Stem and Get Whole Song, you can Stem a Stem.)
Results: Not exactly identical but not very interesting.

Original Song: https://suno.com/song/d3bf9f5f-07bc-4f2a-9815-fab0cd6d84d5
1x and 10x Instrumental clips, shared as wavs. (If you Extend a Stem you can download as wav to workaround the current stem-wav bug)

[runic dock]

Stems are available but far from the expectations.
The challenge was high and suno is not doing better than all the solutions available. Obtaining instrumental stem seems the most chalenging. Suno has taken the path to less degrade the music quality. Than part of vocals remain in instrumental stem.

[digital ruin]

hey there, a few questions:

• what's the best prompts for techno? Suno is really struggling with techno, I get like 1/10 results correct
• how do I ensure to have my vocal preferences? e.g. sometimes I say female vocals but I get male vocals
• is there a way to prompt Suno to use both male and female vocals in specific song sections?
• lastly, do you have any tips on how to ensure I get the genre I want? quite often i don't get what I asked for

that's it for now, relatively new to Suno, will have more questions in the future for sure

[wicked mortar]

Female voice and female singer are options to try also, as for have male and female I did have a song where I had [verse name] and [chorus name] and got a male and female vocal on each part, but I haven’t tested further to see if gender specific names would help

[north niche]

Question ❓. If you like the music and the beat it's generated but want to change the lyrics, is there a way of doing this? Keeping the music but changing the lyrics?

[mystic grail]

2 biggest tips that helped me improve A LOT:

• always use advanced mode and don't use sunos lyrics generator, use chatgpt, remember to tell him to use []
• always add [Intro] at the beginning, and [End] [Silence] at the end, also remember it can do a lot with [] like [laughs] [screams] etc

[warped furnace]

Techno is a very broad term. It's like using EDM. There are so many variants that you would need at least twenty tries to get what you want. Just found out that Suno knows Moombahton. 😁
I would try a specific style like Acid, Minimal Techno, Vocal House or Techno House and so on.
EDM has a lot of sub genres, too. Industrial or Drum and Bass, Dubstep, Electro EDM, Grime, Hardcore ...

Most of the time it works with (fe)male voice or vocals in the style prompts although some styles have their own voice preferences.

Haven't found a sure way. Try sections and generate the parts seperately. Part one with female vocals, then next part male vocals (although it might revert back). It helps having a subscription and more credits to try out until you get what you want.
Have had Suno make duets when I didn't order them. The most likely combination of prompts (in my case) was putting choir, harmony singing and (fe)male vocals in the style section. But sometimes Suno does what it wants. It's not an exact science with AI.

[warped furnace]

Extend the track. Works best if it isn't at the beginning of the song. Just yesterday I extended a song which was repeating verse 2. So I extended after the first time verse 2 was sung. It happens that the glitch "infects" the extension, too. In my case, it helped to extend in the middle of a verse or chorus. Mostly it works for me.

[untold briar]

Overlapping vocals in V3.5 example. Mostly an accident but I think the text formatting: plaintext, (), [] contributed: https://suno.com/song/bdc1f817-bb5c-43ca-8062-853fc9726de5

[brave gull]

!!!!!!

did it take a lot of generations / is it consistent?

[neat lake]

Hi, an accessibility question. Using a screenreader and my song has been chosen for volume 1 of Summer of Suno, but I'm unable to find out which number on the playlist my song is. Would anyone be able to help me out with that please?

[brave gull]

I can do (but idk how to do it)
so if you can give me directions i'd gladly help

[untold briar]

The lyrics prompt was like half the whole poem by Oscar Wild, when it overlapped a little by chance, I Extended from right after a bunch of times until it stuck longer. So not consistent. But maybe with patience?

[neat lake]

If I send you the link to the song would it show?

[untold briar]

I wasn't intentionally trying to overlap, rather just make interesting vocal dynamics, duets, etc. So maybe you can stack on other techniques.

It does make a lot of good duet-ish things consistently though.

[brave gull]

idk maybe someone who actually knows can help
can still send the url in case it does

i'll experiment, i still have 7.8k credits left

[untold briar]

I can help as well. Really just need your Suno username.

[neat lake]

Oh ok, it's: @sentersan2002

[untold briar]

I can see the song, but the playlist is very laggy hard to count the place...

[brave gull]

also, how did you figure out the fact that v3.5 needs -s for styles with spaces? just random experiments?

i got something super scary wtf

demon intro

for a song about couscous

[neat lake]

Yeah I can second that, please let me know if you're able to find out, means really a lot

[untold briar]

I can make a guess, based on scrolling, maybe between 200 and 300?

[unique stone]

Hmm... Was the full song really mono when you downloaded it?

[serene sable]

Hey folks, i have an old song created with v2, is there anyway to process it with the actual version of suno to make it better. I try upload a part with the feature but i can just extend and i lost all the mood and the music it's totally diferent (not an issue it's expected behavior ^^)

[neat lake]

Wow, that would be amazing. Thanks for making the effort anyway, still happy to know I'm somewhere on that list 🙂

[serene sable]

btw i don't understand the purpose of upload if we can't "analyse" it to have something like uploaded part

[unique stone]

Yeah the full song one is mono

[untold briar]

I can't remember who, but another Suno mod first noticed it (I think it wasn't me, at least), and it was obvious enough it stood out with just a few tests on on launch day - for some specific prompts anyway.

[brave gull]

yeah, it's a lot better with dashes

sea-shanty kinda works now

but it still adds electic guitars

[untold briar]

You can repeat with the dash. "sea-shanty, sea-shanty, sea-shanty, indie-shanty, sea-trumpet" (I also like to try compound styles like that)

[brave gull]

right

thanks

[unique stone]

@robust ermine here's the left channel of both clips you posted previously, but with one panned to the right instead.

Weird. Full song sounds mono still. What about the MP3 download for the full song?

[robust ermine]

yeah that's def mono, i don't know why though. Here's what I get with reaper, and you can see that the waveforms are different too

channels don't cross-bleed, I double-checked by muting one

[unique stone]

Yours seems to be the left and right channels. The top waveform has a quiet part later on, which is in the left channel.

[mild rain]

ohh excited just read there is now Stem separation, anyone tried this yet? any good?

[robust ermine]

hm, but that's not what panning does. Okay let me try and do it your way with original lefts to left and right

[unique stone]

I suspect what's happened is the panning has just quietened the opposite channel but left it mixed as-is.

Here's what it looks like in audacity when I take the left channel only

[robust ermine]

damn, you're right. I'm now starting to wonder if I rendered the same model. Let me try sharing the original links

v3.5: https://suno.com/song/0649c274-bda4-4111-bcb1-eb1fe23ef871
v3: https://suno.com/song/13bcd4d1-c11f-4358-8916-5d7d1c2ae9b4

@unique stone same. Man, you're totally right. I can't believe mind can play these tricks on me lol

[mild rain]

hmm its ok just tried it, sounds awful.... on well I'll just add this to all the other Suno backward steps...

[unique stone]

Seems to work well for vocals. Not so well for instrumental. In the brief testing I gave it. Is it really a backwards step if it's a feature that Suno didn't even offer before?

[unique stone]

Just to make sure, I took iiCaptain's suggestion from general chat and inverted one waveform over the other and it was totally silent so they are 1:1 identical. I was curious in case there was a bug 🙂

[robust ermine]

yeah, good trick too

[unique stone]

That one sounds stereo!

[mild rain]

plenty of offline options can do this I was hoping being that it was suno generated it would be a clean separation, I already have serious audio quality issues which I have demonstrated time and time again this makes a bad sounding creation just worse for me.

[unique stone]

It should improve (it's just v1 of the feature.) It's useful to be able to isolate a Suno vocal to potentially reuse it later. Admittedly I'll probably continue using external tools to split things since I don't care too much about keeping everything within Suno (plus I don't want to have even more stuff in my library 😄 )

Hmm so it's OK now? When did you last download the WAV?

[fickle siren]

Last night was another really good lesson in how important lyric structure, especially in the chorus is.

I was banging my head against a wall trying to get a song to be more upbeat, and realized that some of my lyrics, especially in the lyrics, were just trying to do too much on one line. Breaking them up a bit seemed to give Suno more freedom to try different melodies and it ended up unlocking the sound I was looking for.

[unique stone]

So it's still bad? (sorry, getting mixed up here 😄 )

If you use "get whole song" on it again and download the new song it creates, as a wav, does it still sound the same?

[mild rain]

Thats actually a good point each stem separation adds another 2 items to the create page.... really hope at some point they will allow us to move stuff out of this page in to custom folders. Pages and Pages is so time consuming.

[unique stone]

The earlier ones you were posting were all mono so I expect it should sound closer to the mp3 (which was stereo)

[crisp fractal]

Hi.🖖 I generated a text and inserted an alternating command after each sentence, style [spoken word], [singing], [male voice], [female voice], etc. 😎.
But he hardly obeys. Does anyone know the correct way to generate these commands so that they work in Suno? Thanks. 🙏

[quiet prairie]

Anyone here use Zynaptiq Unchirp? Trying to find good default settings for Suno chirping.

[runic dock]

Nothing new except that they are trying to remove less vocals to get instrumentals. The result is vocals remaining in instrumentals. Nothing really better.

[dim forum]

exactly , it has phasing issues its like u use any online seperator , and point of the matter is not the accapella...scrw the accapella
Its the instrumental stems so u can clearly hear whats in there and use some parts and other resample them in order to release it clean and pro sounding for labels

i appreciate the effort ofc...but this way its kinda unusable tbh...

the only way it could be better would be if you could generate dry version no effects , reverb , delay etc that way it could be a bit usable with stemm splitter

[runic dock]

This version is not intended for labels, but for karaoke in bars.

[dim forum]

where does it say that ?

[runic dock]

Its a joke

[dim forum]

lol

[untold briar]

This is the song I was working on when I forked off to focus on that overlapping clip. The () [] line formatting still does interesting things to the vocals, I think.
https://suno.com/song/06a23e96-9b75-443e-ba04-526fc3cc1a31

[nimble smelt]

When will the app be optimized for iPad? Thx

[rose jasper]

On a note for the latest update btw Devs, it seems as though, if you include quotes like: aaaaaa-a-a-aaa <- These do NOT count towards the vocals stems

Instead these are added into "Instruments"

[zealous venture]

I was hoping stems would include drums, keys, bass, vocals, guitars etc not just the audio seeming like it was sent to a stem splitter website and still has a ton of artifacts and the vocal reverb is still I the instrumental, I appreciate them listening to our requests but stems feature is in usable TBH

[dim forum]

Please dont post tracks here

[mild rain]

I said exactly the same. I was hoping as it was Suno it would be a clean separation but was totally gutted when it was just a fuzzy artefact filled mess...... Not sure what the benefit is here from any of the third party offerings people have been using, and as someone else mentioned this also adds two extra entries in to the create screen as if this wasn't busy enough already with its pages and pages and pages and pages and pages and pages and pages and pages and pages and pages 😋

[naive quail]

tracks are regularly posted here, usually to highlight different techniques and such.

[charred merlin]

#📱┃bugs-ios would be a better place, as this is more for questions about how to use suno itself ❤️

[sturdy flax]

Agreed, I want to use it for collaboration on songs by several IRL artists but at this quality one can really only use it for writings and ideation

[naive quail]

...and maybe karaoke night 😄

[meager crescent]

Thank you for the STEMs but they do not line up with original see https://suno.com/song/6df5f0e5-ad6a-493a-8a85-75d25f62ed4b is out of line with its stems https://suno.com/song/19ec0061-7cb2-4d21-9004-7c37bfcb2a28 and https://suno.com/song/59d943ce-686f-49f6-a8f5-5f8507005e2f. Not by much, but enough to hears.

[meager crescent]

This is instrumental song https://suno.com/song/59d943ce-686f-49f6-a8f5-5f8507005e2f

[amber vigil]

ah i see the effect now as more desirable. When you posted in suno showcase, i was on the go and hadn’t been able to listen. This is unbelievably cool.

The big question is always in replicatability with Suno. I might play around this coming weekend when i have time.

[teal raft]

iicaptain uploaded a reversed sample and Suno carried it through the whole song

[amber vigil]

oooh

[teal raft]

Lyrics needed to be phonetically spelled out

But the music stayed with the reversed style

[amber vigil]

we can finally do backward masking for secret satanic messages? (Joke)

when i was a kid, parents would routinely wring their hands and wonder about secret messages in songs. It was quite a preoccupation for way too many.

[teal raft]

Satanic Panic

[crisp fractal]

Why doesn't this work? https://sunoaiwiki.com/resources/2024-05-13-list-of-metatags/

[rose jasper]

Do you have any examples?

[toxic solar]

Hi! can someone explain to me what happens if i upload an audio of an already suno made song, and ask it to have different lyrics? thanks!

[charred merlin]

[kevin burps into the microphone]

[crisp fractal]

he confuses and mixes up speaking when he should be singing rap, he uses a female voice when it should be male, and so on.

[hollow hare]

Most of those tags were taken from the old Suno v1 metatag FAQ. As a side note, the unofficial wiki we like to use is https://suno.wiki which is more generalized (though Suno does have updating their official FAQ/wiki on the list of things to do down the road).

Mind you, these tags are like suggestions to the AI and it would perform them if it makes sense at that point in the song- assuming the tag you are using is not outdated (a few no longer work, or at least did not work when I tried them)

Did you use a rap genre in the style prompt? Make it one of the first few keywords for priority. Note that using commas makes the entries treated as a set of keywords rather than individual (see #💬┃general-chat message for what I mean).

Sometimes you need to add more weight to get what you want, especially depending on the genre. Repeating the keyword will do that, e.g. female voice, clear female vocals, woman, ...

[toxic solar]

someone? :´ )

[hollow hare]

It sounds like trying to do an Extend with more steps

[carmine osprey]

The anouncmenet said Pro and Premier users -- what is a premier user?

[hollow hare]

Pro is the $10/month plan whereas premier is the $30/month plan. Same terms with more credits

(2 500 credits for Pro, 10 000 for premier per month)

[carmine osprey]

ah thank you!

[hollow hare]

You're welcome

[naive quail]

A regular occurrence... 😡

[toxic solar]

i mean, what i wanna do is change some lyrics of the original audio, is that possible?

cause i really liked the style that suno took on the take

is there any way to take an audio and make some variants with the same original melody?

or al least a similar one

[hollow hare]

Not really since we cannot control the seed, though I did not test for that specifically.. You still have a good chance of the song changing the shorter the clip/extension time is.

You should be able to get something similar to the same melody but the odds of getting the exact same one is very low

[rose jasper]

The best thing to do, find one you like, and start to Extend from the point in where you would like to keep the style at @toxic solar

[crisp fractal]

I replaced the old tags with the new ones, but the mystery remains. Unfortunately, Suno has a strong personality and does whatever he sees fit. and the credits are gone… 😕

[hollow hare]

Unfortunately that is AI for you, have to play around and learn how to influence it to get what you want.

Do you mind sharing the link with me? I want to take a shot (just let me know what you are intending). You can DM me it if you do not want anyone else to see it.

If you are on the free plan, the credits will refresh in about 8.5 hrs

[flint haven]

Free Credits Refresh

Refreshes at 0300 UTC+0 (<t:1713236400:t> in your local time)

[magic gazelle]

Suno...stop being lazy. I can literally most of the time type in on my songs "Piano only" and after 45 tries till generate the piano part of the song with no artifacts and nothing else. Why would half step on the stem separation? It doesn't need to be phase isolation or whatever it is. Just allow us to type in the prompts for what we want out of the song and let it generate it and then bam. Or let it actually separate the parts individually.

[charred merlin]

stop being lazy

[sinful merlin]

Yo, I vomited a helpful newbie guide to music production and will post it soon in the subreddit

[charred merlin]

Have you tried re-approaching the problem from another angle? There's some really bright minds in this community that have found remarkable ways to use Suno, given the different features available (extend, upload, vocal-stem splitting, etc)

[charred merlin]

YO! You rock 🤘

Is it something you could post https://discord.com/channels/1069381916492562582/1252267444773326848 here as well? ❤️

[charred merlin]

NICE! I know it's called "newbies guide", but do you think it caters to the casual-to-intermediate crowd suno crowd in general, or is this specifically for people who have been finding success with Suno and want to take their production to the next level, and start to "legitimize" the production process to a degree?

[worthy phoenix]

Who knows how I can do a duet properly? I tried to do a duet, but anyway it came out good in this https://suno.com/song/cc2a9a96-ec52-4c91-9fb9-960085c9c1a7

[magic gazelle]

I've tried a couple different approaches but I havent found one that works without me having to utilize a whole bunch of credits on 2 minute loops of nothing but silence or anything like that. I can get enough pieces of the song to put it together sometimes but it really just depends honestly. I think they should just allow you to type in "Guitar only" Or "Vocal only". And then it give you that specific part or Percussion only. Then bam. I wont have a problem spending my credits for that.

[toxic solar]

ooohh oke thank u so much @rose jasper @hollow hare

[rare oak]

Heyo, is it possible to retrieve the prompts used to generate the tracks images ?

[unkempt shale]

Tried the STEMS. Are they just using the L/R invert technique to pull or AI? I am hearing vocals in the instrumental and hard gating in the vocals.

On another note, the EXTEND feature is still not extending songs properly. I try to extend a 2:12 song at the 2:00 mark (way before a fadeout or anything) and I'm only getting :10-:15 seconds of an extension. I need 3-4 minutes of extension.

Would be nice if during the Extend I have a checkbox "Use the whole 4 minutes".

[runic dock]

Udio has released today, stems download with bass, drums, others and vocals.
Suno has released stems today too, with only 2 tracks vocals and instrumentals.
Curious !!!!!

[hollow hare]

Most likely a coincidence, not like both company planned the release together.
Regardless, this is not the channel to talk about that, more of an #🗣️┃chill-chat

[charred merlin]

Not that I've ever seen discussed, at least.

[charred merlin]

Do you have any technical questions or detailed feedback on the matter, or are you just excited to gossip?

[runic dock]

Detailed feedback to devs already provided using the google form. Technical feedback provided here too #📚┃suno-school message
Maybe not enougth?

[rapid pulsar]

Wow, does Suno really only support up to 60 second uploads? That's a deal breaker. Guess I'm sticking with Udio.

[charred merlin]

The earlier comment you linked above is helpful feedback -- however the comment that Paimon responded to doesn't offer itself as a question, it's just more of a statement that a different company offers a different service.

I'm not trying to be a jerk, btw, it's absolutely something worth talking about if there's other tools to play with, but I wasn't sure if there was a technical follow-up that you wanted to discuss about the difference in the two, or if it was wanting to chat about it (which again it totally cool, but could be directed to #🗣️┃chill-chat like Paimon said just to keep the tech-chat "tidy" ❤️ )

@rapid pulsar, yeah, 6-60s as of the current implementation -- they're still squeezing really cool stuff out of the V3x series, and I have literally no idea what to expect next 🤣 Kinda like OpenAI is breaking out GPT4, GPT4 turbo, GPT4o, and now GPT4 mini -- maybe they've discovered something really cool that they can build off with the existing models, maybe they're taming the wild-unicorn that v4 must be -- I'm just stoked it's happening so frequently, with such transparency (https://discord.com/channels/1069381916492562582/1163842729231528047), and we don't need to wait AGES between super soulless-feeling corporate bullshit -- they just talk to us

[rapid pulsar]

Also, I have no idea what happened to the UI since I last used it. I can't scroll anything on desktop and navigation is very difficult. I have to zoom out my browser by more than 50% to fill in all forms.

[charred merlin]

#🪲┃bugs-web might be a more targeted place for assistance regarding that, if it's a reproducible issue?

[rapid pulsar]

I used to be pretty loyal to ChatGPT but completely switched to Claude 3.5 when they overpromised and underdelivered. I wish Suno offered enough to bring me back. I heard it's now more possible to "remix" your songs. But almost all of my songs are longer than 60 seconds so it's useless to me. Udio can do it, even if it's slightly fiddly to get it to work how I want.

Still, glad to hear that Suno's communication is good. Maybe if they fix this deal breaker by 4.0 I will indeed return.

[rapid pulsar]

Yeah, you're probably right about that.

[grand stream]

Hey, I'm having trouble logging in to Suno AI using my mobile browser. When I hit the sign-in button, it just spins forever. Even after clearing cookies and cached data, the page is still loading.

[hollow hare]

If you look closely, you will see that the Cloudflare validation failed

[rapid pulsar]

I'm very confused by the audio upload in Suno. I have chosen to use a 60-second audio upload. I chose instrumental and a desired style for the output. But the results sound nothing like the original song. It's almost as if it never paid attention to the upload and just did text-to-audio. Quite baffling.

What makes it worse is how after upload, there's no indication in the interface that it's using the uploaded audio, which adds to the ambiguity. Udio on the other hand is clear and straightforward about what you've uploaded.

[grand stream]

I understand, but I've been dealing with this issue for more than a day.

[unique stone]

When you've uploaded the audio, it appears in your library. You need to press the "extend" button on that clip to use it.

[rapid pulsar]

I thought it had a remix function.

[weary heath]

how can i make vocals from deep to high for a remix?

Example:

Chorus (normal Voice)
With u i can be free

Chorus 2 (from deep to high pitch Vocals)
be free, be free, be free, be free

[hollow hare]

Can't really help you with it personally aside from some generic advice, seems to be something you may want to contact Cloudflare about specifically.

1. Did you click the Having Troubles link and reading their entry and following their recommendations?

2. Did you try using a search engine to see if anyone else encountered this issue and had gotten it fixed?

3. Did you try with another browser in case there is an issue with any installed extensions or settings you have (e.g. disabled all third party cookies or had javascript disabled)?

4. If checking another browser failed, did you try using a VPN to ensure it is not a DNS associated issue?

Well, truthfully, #4 is very unlikely. I think #3 is most likely the culprit, especially if you did some rather extensive privacy measures (which is what I did as well but often left the content breaking stuff alone)

[crisp tinsel]

Is there a way to end the song when I want it to end? Sometime Suno forces a last extra repeat of a verse and just before the last few seconds of the 2min mark and gets cut prematurely.

[weary heath]

if u have a idea please tell me 🙂

[zealous summit]

On the website, is there supposed to be some kind of player that pops up playing song?

It doesn't for me

[runic dock]

Just to be clear, Suno is my preferred music AI tool, but the stems feature seems to have been released too early imho. The quality isn't good enough and falls short of expectations. I was just wondering if the feature was released due to competition pressure. So the intention was not to promote compétition but to say we could wait a little bit to have better quality. But yeah i aggree this kind of discussion can go in off topic or general suno.

[zealous summit]

I can play songs by clicking on the icon, but it's hard to tell where to extend from if I can't see the time

[rigid ruin]

I am dealing with a website/account issue that is preventing me from creating songs? I reached out to billing@suno, but it has been several days of no response. I have tried everything possible on my end.

[raven turtle]

Agreed. Getting a lot of vocal bleed in the instrumental mix stem. Feels like a bit of a rush release..I'm.still sticking with Fadr for now on this..Stems are just not clean enough on Suno's end ATM.

[charred merlin]

You rock, I was worried that I would have come across as discouraging you from sharing your opinions ❤️ It's interesting to hear about what different tools are being released, and I really enjoy seeing all of the cool stuff that our community's creatives share -- like Hedra, Runway, Websim, Udio, etc ❤️

Admittedly I've only used the stem splitting tools from UVR, OpenVino, and BandLab -- I haven't had opportunities to really dive in the differences between them beyond the fact that I used Audacity + OpenVino to mix & master my first album, BandLab for work near the end of the album, only first impressions of others.

Out of curiosity, what splitters have you tested?

[heady sun]

It will get fixed. I reached out to billing and they fixed my account. Had the same issue where I couldn't create songs too.

[charred merlin]

I'd want someone else to back me up before I said it with any sort of confidence, but I think email support is active from 9-5 EST, Monday - Friday, if it helps shed any light on the expected turnaround times?

If it's been a number of days, it can't hurt to reclarify, and "reintroduce" the problem -- I've found that sharing an overview of the issue from your perspective can help prompt a more direct response ❤️

[zealous summit]

ty, probably my web browser

[austere dragon]

which works better [key change] or [key shift]?

[vale isle]

Why do so many of my songs glitch and repeat several times randomly?

It’s happening more and more

[hollow hare]

See this great tip by Shiroken
https://discord.com/channels/1069381916492562582/1261146873331318835

[foggy leaf]

i can still hear vocals in my stemps

[weary wharf]

From my experience [key change] works better
But! You mostly need a [bridge] before that. Else the probability for Suno doing a key change drops a lot.

Something like this works for me 70-80%:

[chorus]
... lyrics

[bridge]
... lyrics

[key change]
[chorus x2]
... lyrics

[supple trail]

Pois é, Tribuzeus!!! Tentei fazer música para uma letra minha em inglês hoje e erros grosseiros estão acontecendo. Até musica com Female quando eu coloquei claramente MALE. Fiz umas 10 vezes e não consegui nada que prestasse. Antes, das 6 músicas que já fiz, a maioria foi de primeira. E todas de boa qualidade. Pior é que mandei email para o billing@worthy grotto e nem eles sabem como resolver os problemas que criaram com as mudanças ocorridas depois de 01/07/24. É uma pena... era tão bom.

[austere dragon]

This is mostly what I do. I even successfully put a key change tag right before the last word in a bridge and it worked! Every other time though it would just sing “key change” as part of the lyrics.

[weary wharf]

Too many tags bleed into song in the lyrics. Try to remove the ones you do not need.

[a][b][c] <- will probably start bleeding if you are doing this a lot.

[sinful merlin]

There was another guide posted today that was more of a everything you ever need to know about suno that it's more of a newbie guide yeah

[scarlet verge]

Hey guys, I'd like some help on what to do.
I'm having problems with my credits, I'm not getting my free credits, they're not coming to me, I only have 75 credits and the free ones aren't coming back, whether I use them or not, whether I use a few or all 50 credits. What should I do?

[charred merlin]

/refresh

[flint haven]

Free Credits Refresh

Refreshes at 0300 UTC+0 (<t:1713236400:t> in your local time)

[charred merlin]

@scarlet verge if you're on a free plan, your "allowance" of daily subscription-credits resets back to 50 at the time listed above -- if you already have 50 subscription-credits, it doesn't stockpile any of them, you just remain at 50. It looks like the credits should have reset 14 minutes ago?

[scarlet verge]

Thank you 🙂

[charred merlin]

You're very welcome, my dude -- I swear they need to add a "credits history" popup on the https://www.suno.com/account page just so people can get a better idea of when their credits come in and out of the account -- it would help to better inform decisions regarding what the best subscription plan would be, or to provide additional context before purchasing extra credits (ie. do I NEED to make this song now? Can I wait until the next subscription cycle rolls over, whether daily or monthly? Am I wasting credits by subscribing to too high of a tier? etc.)

[queen flame]

/refresh

[flint haven]

Free Credits Refresh

Refreshes at 0300 UTC+0 (<t:1713236400:t> in your local time)

[queen flame]

/credit

[spare stump]

@charred merlin i'm not sure what others are saying but my PC is lagging big time with the new UI scroll feature. I prefer the previous version we had up until yesterday. It lags very badly and I only have a 6 month old PC

[shrewd garden]

Hi, i am having changing choruses in my song. Every chorus is a little different from the previous.
Suno unfortunately keeps repeating the first chorus. Is there a way to force a rerender of each chorus individually?

[true night]

i know this is a very generic question but have you tried diffrent browsers, disabled any extensions on youre browser?

[true night]

have you tried adding lyric manualy like the verses then chours , then make song then add new verses and then the tiny bit diffrent chours and press extend song and do it that way? with same chours? coming *

[weary heath]

Do you have and idea?

[runic dock]

I've tried UVR, Moises, and FL Studio, and I've noticed a common issue across all these tools. Music is composed of a complex mix of frequencies and overlapping sound waves from various instruments and vocals. This complexity makes accurate separation challenging because the individual components are not easily distinguishable.

[true pier]

I am more accepting of the last version. This version =_= is so difficult to operate.

https://tenor.com/view/cat-stressed-nervous-awkward-uncomfortable-gif-27051002

[shrewd garden]

I think to remember, that in previous projects, that it still used the first chorus.
But I'll try.

[true night]

idk if it works but thats how i tried before to go forth on the prev projects with diffrent chourses buts its a bit 50\50 hit and miss tbh..

[mighty scroll]

Artifacting gets noticeably worse at 90-120 seconds

and then takes another big dive on any extension

Artifacting is substantially reduced when extending from an original work if that work is cleanly recorded and well mastered

stems are currently unusable for anything near as I can tell

dolby atmos cleans up the vocals 20-25% which is better thna bandlab

but expensive

Writing 6 lines of lyrics of equal metric feet gets you 8 bars in a xxxxx-x- pattern very consistently, close to 90% of the time, while writing 4 lines of lyrics gets you the following patterns in about equal frequency: xxxx xxx-x xx-xx- xx-xx so you really only get 4 bars about 20% of the time

this is what I've learned generating 600 songs in the last 18 days. DAW to suno to DAW to mastering suite takes ~20 songs to generate 1 good track, I have 32 results I'm happy with. However public at large, no matter how much you rework the original material, invariably comments "why do u music sound lik it thru a phone fam lame breh" which I suppose speaks to the effete sophistication of genpop when it comes to audio quality. 30 ys ago fraunhofer cbr 120 mp3s made to ISO/IEC 13818-3 spec were state of the art and we were amazed, suno's output is about on par with that quality. So we have maybe 30 years of catching up to do

[mystic flume]

You can get some good results by using them to augment the original.

[mighty scroll]

luckily ai jumping ahead 30 years doesn't seem likely to take 30 years

[mystic flume]

Beefing up the bass and drums really thickens the song so the highs and vocals don't sound so tinny (no relation).

But yeah, on their own the stems are pretty lackluster no matter what you use to pull them, paid or free all sounds about the same (because I'm sure most of them are using UVR).

[mighty scroll]

I guess I should clarify, suno's native stem pulling is unusuable

pulling stem using other tools is helpful

[mystic flume]

Oh huh, I didn't see the news. Interesting.

[mighty scroll]

dolby atmos does a great job of isolating and tempering nearly all the garbled high end stuff

[mystic flume]

For $9/mo

Ah, it's just vocal / instrumental. I dunno, this first one I tried sounds pretty good. But going back to grab stems on songs I've already edited, remixed, and mastered sounds like woof. I'm gonna keep it in mind for the next one, so I'll post my thoughts then.

[mighty scroll]

^ Don't waste your time, yet

the splits get worse as the song goes on

more instrumental in the vocal and vice versa

and more artifacting than the source in both

[unique stone]

Suno's stems are just "version 1" so room for improvement in future. It's good for keeping things within Suno but if you're working with stuff in a DAW external tools still give the best results (I tried some online ones then found UVR and that does what I need.) Other AI services can help with improving or swapping the vocals for higher quality ones. For instruments I tend to just overlay similar sounding instruments and maybe keep a bit of the original audio for warmth.

[mystic flume]

Well, the songs themselves get worse as they go on, right? So that could be why the stems sound worse. But at first blush it looks like they're running them through a regular UVR style tool. My hope is there will eventually be a way to generate them as stems.

[mystic flume]

Improving the vocals? Tell me more.

[unique stone]

True stems will inevitably come in future, but I feel it's a long way off due to the added complexity of processing and the need for lots more training data (such as having true stems to begin with)

[mighty scroll]

yeah, I start the song in DAW, pull 3-4 chunks into a 60 second vignette, run it through suno, take the resulting 3 minutes and dub it back into the original track and that seems best so far

suno gets WILD with tempo changes tho

I've had songs speed up as much as 250%

(extended twice)

[mystic flume]

So you're attempting to re-generate the instrumental without lyrics?

[mighty scroll]

I'm attempting to take an existing, complete piece of music with full instrumentation and spare/no vocals and fill it out with additional vocals and solos (suno seems to nail saxophone and harmonica solos for example)

[unique stone]

I've experimented a little with lalal.ai kits.ai and audimee for replacing the vocals. By far the best improvement I've found for vocals is to upload the Suno song into another AI music generator (yes, that one...) and tell it to "remix" with minimal changes, which essentially regenerates the audio but using their model. I've had mixed results and the instrumental can get degraded by doing it but the vocals tend to end up nice. I haven't done anything with the output yet though - still experimenting. I expect stem splitting and putting the improved vocal in place of the original would work quite well.

I hope V4 brings better vocals to make all of this unnecessary but I quite enjoy chucking bits of audio around into different tools 😆

[mystic flume]

Ugh, I hate that other one; the interface feels so obtuse to me. Like I'm staring at the console of the space shuttle. I just filled out the survey Suno linked in the What's New? section (https://forms.gle/xWmRmzHQrn8Bmrzu9) and gave them some feedback on what I hope for from the stems features. Might be useful for the rest of us to chime in too.

[mystic flume]

I never really did anything with audio before discovering Suno, but I've found it to be fascinatingly fulfilling.

Now I'm talking about stems and DAWs and doing a whole production in Audition. It's quite thrilling indeed.

[unique stone]

Yeah Suno is a lot more straightforward to use. Plus it rarely gives a "bad" song (just, overall audio quality is limited.) Being able to start listening soon after pressing the Create button is so useful as well. I've got a lot of ideas from it to work on but I start experimenting and then get more ideas. Over the weekend I regenerated the same piece of song dozens of times just because I wanted to hear alternatives 😆

The day that true stems are available, I'll be very very happy. For now, I'll work with what's available.

[mighty scroll]

I think it's neat, but I can't get anyone else on board. If the quality of output didn't offend their princess-and-the-pea ears I think I could get more people excited about it

[charred merlin]

Definition of a “Sunovator” — stamped, certified, certification is in the mail ❤️

Really though, so many people are in the exact same position, but might feel like they don’t know how or where to take those next steps.

I appreciate the feedback regarding LANDR (just realized last night the name means “left and right) — gonna be REALLY exploring it over the coming days, so I’ll keep in touch re: any tools or on-site learning content you found helpful 💜

[charred merlin]

Regardless how easy it is for us, it has to be absolutely mind blowing for the New-to-casual users who never would have made it beyond the friction point of taking songs to another platform for further production.

(Note to self- during:days after stem release in:genchat search for excited keywords, follow up with excited, approachable, or eloquent users as to continued thoughts and expected use cases)

i thought subscript was allowed in here , but woodhouse yelled that i was trying to make the words big.

I was doing the small words thing, @flint haven 🤦‍♂️🙈

[unique stone]

If it's too small he can't read them. Woodhouse is very grumpy about these things.

[unique stone]

It's mind-blowing even for me 😆

Even after using it for several months

[charred merlin]

I understand — he’s senile and currently doesn’t know me from Adam.

Give him some sugared grapefruit, a firm and lasting hug, and let him know that the weird guy with long hair still thinks very fondly of him.

[charred merlin]

I’m still discovering on a near-daily basis just how different my musical life is…

And music is a big thing for me. 👀

Legitimately never been more excited for something in my life.

I think I used to play or make video games or something? I swear there were other things I used to care about beyond cats and Suno.

[mighty scroll]

https://suno.com/song/def2f1a7-d292-4832-895c-9cafba5f86be I think maybe I'm too tired. I laughed at this until i hurt my shoulder

I think maybe some of my troubles with audio quality are sourced by my lyric choice?

[shrewd garden]

I have tried a different approach.
My chorusses now have an index like [chorus 1], [chorus 2], etc.
The Songs I have done by now seem to respect the index and do generate new chorus

[true night]

so that worked then? im happy to hear it got sorted in a way that works for you , but also appricate the feedback that it got sorted 🙂

[cold kiln]

How do I get SUNO to make an abrupt transition to a heavier tone in the music? No matter what prompt I write and no matter what I do, it doesn't want to do a transition of the music.
I have a song that starts with classical guitar and I can't add rock elements to the song.
All I've been able to do is add low bass but that's not what I want to do.
My goal is to make a development in a rock direction like in My Immortal by Evanescence at the end of the song.
HELP please!

[charred merlin]

#📚┃suno-school message

[charred merlin]

https://suno.com/song/65a2e520-e54a-4def-be67-d483a08a49eb

100%, creating a song with no provided style prompt, and just the lyrics “I’m so angry right now” results in a metal song.

[mighty scroll]

compare the clarity of the lyrics in the second verse (high clarity) versus the lower clarity of the first and third verses tho: https://suno.com/song/def2f1a7-d292-4832-895c-9cafba5f86be

[naive quail]

This is using a method that's pretty similar to what @charred merlin linked above. The initial switch to metal is more abrupt than I'd prefer, but I didn't want to keep throwing credits at it to get something smoother. https://suno.com/song/18320a68-73cc-44d2-b8b5-5ae528b6b8fe

wtf is that squeak at 00:50? I can't tell w/o my headphones/iems

[charred merlin]

Did I ever ask you for brutal critique on the terminator-extension thing? I apologize if you shared and I missed it

[naive quail]

It looks good. I think of the 1s extensions as pivot points. I used one to switch to metal, merged the parts via Get Whole Song and then immediately used a second one after the metal kicked in so that it'd keep the heavy guitar tone while having full access to the earlier classical guitar for reference.

(if that makes sense... 'haven't done my espresso yet; I'm in Zombie mode)

[charred merlin]

I had a moment of pause trying to think of "names" for the sections, and while I agree that 1s-extension IS the point that you pivot on, I felt that without accentuating the fact that there's supposed to be "a kill gap" between the tracks, calling part 2 the terminator helped suggest that part 3 is supposed to be "the style we're pivoting toward", but didn't want anyone to get confused by the naming and think they needed to add style prompts to part 2

[charred merlin]

I forget that I'm an hour ahead of you -- you absolutely don't need to rush and throw a thumbs up or a reply to the random shit I ping you with at 6am your time 🤣 😛 -- always appreciative when you're free, my dude

https://tenor.com/view/tired-confused-wake-up-squinting-gif-4201426486566488898

[naive quail]

can you tell what what 00:50 squeak is? Is it just a stray note, or a mouse getting stomped on?

[charred merlin]

Howard DeWitt Hooper III at 6am trying to make sense of wtf I just sent him

[naive quail]

https://suno.com/song/a628210c-795e-48db-b4e1-de9ae5102108

[charred merlin]

oh man, that pivot at 0:23 is SO good, just a solid, classic "walk up"

The squeak sounds like it diffused a string bend/slide into the flamenco acoustic? 100% what it is

[naive quail]

'was really hoping that it was a mouse. That mouse is Dun!

[shrewd garden]

Well, it doesn't work always. But most of the times.

[charred merlin]

@compact marlin the way that I described it to my mom the other day when she had no idea what the heck Suno was --

Think about police computers back in the day -- they took known photographs, added blur or morph effects, then trained a computer to understand how to reverse those blurry pictures back into a face we recognize.

After training a LOT like this, eventually they could make a purely random blur image and the computer could "imagine" what that person could look like, even though that face likely doesn't exist in the real world.

After the RLHF (reinforced learning from human feedback) stage, you've usually guided the model to the point where you can try to recall specific information from the training data by if it was included (in this case, eye colour, hair style, freckles, etc).

In regards to the audio equivalent, it could work very similarly to the static you hear on radio or TV -- they could have taken training data, and progressively added audio-noise, then trained a model to take similar "reverse-engineering steps" to translate the static into a specifically guided audio style.

The important thing to keep in mind is that it's all literally science fiction magic, and if it doesn't impress you yet, wait a month for the ridiculously comptitive market to push each other to greatness 😅

[cold kiln]

God, how to END the song? its forever generate more music but i need to just END the song, any comand for this???
[End] or [Outro] didnt work

OMG, - 200 credits just trying to end this madness

[compact marlin]

what i have used that word well are [End], [Fade Out] and [Fade to End]

but it really depends how you really want to finish it

from suno.wiki: "[Big Finish] may change the melody or tempo to create a climax." but I really didn't like, seems that the AI dont interprete that properly, but maybe worth trying?

[charred merlin]

#📚┃suno-school message

ONLY use the word "end" for the style prompt in part 3 / the pivot

https://tenor.com/view/bart-say-the-line-gif-22118591

"You don't code/command Suno, you talk to it and give it suggestions."

Not to say that stuff like [end] doesn't work, but if you look back through all of the #😜┃suno-memes you'll notice a trend that people feel you need to over-repeat that metatag/style in order to have it be respected, but that's just skewing the weights of the other metatags & prompts.

All adding [outro] or [end] or [fade out] does is suggest to Suno that the continuation of the song could try to sway toward clips of music that have those "keywords" associated with their training data -- you're not telling Suno "stop making music at this point", you're telling it "remember how some songs have a cool outro part? Maybe you could take influence from what you understand about music from those songs?"

[cold kiln]

This didnt work for me....
I was able to do a rock transition with this guide, thank you for that, however I can't finish the song in this

[charred merlin]

If you'd like, I'd be happy to assist in DMs?

I'm not "the originator" of this idea, as it's just a very powerful way to use the "Extend" function to effectively kill any musical context from the earlier section

However, I "discovered" this method as a way to end my songs whenever I'd like, and only later used them to change styles

For an example, I wanted to make a song for my post in suno-memes, but it "would be funnier" as a really short song, so I stopped it around 30 seconds

[charred merlin]

I know I speak very flowery and colourfully -- but I am worried it makes my lessons hard to understand 😦 -- I would really value your friendship if you were willing to work collaboratively with me? I help make the song you want, and you tell me what was most confusing?

❤️ 🙏

Not just for me, I believe this feedback could help me help others

[stiff sable]

this makes me wonder about the potential idea of suno teaming up with music libraries in the future for training data. aside from genuine instrumental versions alongside full versions, most provide stems as well (like Epidemic Sound for example having a drums, bass, vocals, and instruments stem for every song) which could be very valuable for training stem-splitting models in the future.
in the past couple of years too, most of these libraries have moved away from the 'traditional, cheesy' royalty-free music and gone for a mainstream, streaming-world approach, signing genuine artists, which would probably help make data trained from it more 'usable' too.

no idea about logistics, this is just a wacky concept i thought of, but smthn like that could be a fine solution for gathering enough training data for these requirements in future models, if deals were to be worked out with these companies to provide their music

[charred merlin]

LANDR's Fair Trade AI program is trying to solve this kinda thing -- by letting artists choose to opt-in, they can get a "piece of the pie" when their songs are used in training data -- plus you'd be able to include metadata about your tracks to ACTUALLY allow future models to let you specify what kind of BPM you want to use.

Starting to actually understand this whole web3 music thing -- as of right now the only actionably-implementable use case is "reinventing" payment kinda stuff, so that way it's harder for managers to hide profits from artists because it's all on a ledger, and while distilling it to just saying "web3 makes it easier for you to get paid" is super clickbaity, it gets the current idea across.

What I'm excited about is the possibility that MORE information could be provided in the training datasets, whether or not that data is used consistently by each service that accesses the data, but it sort of provides a song-side API for developing services to consider and build relevant products around.

A lot of the complaints about what generative AI DOESN'T do seem to end the conversation at that point, but I'm curious about knowing WHERE things are moving.

I'm not really sure of any other companies that are currently tackling that sort of thing, and while it's not intended to be free and open-sourced in the same way, I've been told on good authority that it's not unreasonable to compare what LANDR is attempting to do with their "Fair Trade AI" program to a business-facing version of how LAION attempted to tackle the issue of safety and guidance in training data for image sets.

re: LANDR@mystic flume 💜
re: what web3 could actually mean for music in the future? @sturdy flax 🤘

[stiff sable]

ohh yeah i was looking into the LANDR Fair Traide program a couple of days ago bcos i thought it looked pretty interesting but you can currently only be a part of it with music you actually distribute through LANDR.... which makes sense... but i dont want to use LANDR distro so thats off the table for now

[golden abyss]

the prompts and "lyrics" are a good example of "structure instructions" for instrumental music without using the instrumental "tick"

https://suno.com/song/be2c34d9-9775-43bf-84d6-95c480596c5f

[hot trout]

The ability to separate out stems is great, but it runs into the same issue that every utility I have ever found to do that job runs into: the instrumental tracks get awfully muddy anywhere that lyrics were formerly overlaid. Rather than having stem separation be an option available to Pro and Premiere users, I feel like songs in Suno should be GENERATED as separate stems, and the option to download said stems should be offered to paid users. It would eliminate that problem completely, and I would happily burn extra credits just to have fully functional audio stems I could edit and mix in other software. Hell, you could even ask me to spend credits for each separate instrumental layer plus the vocals, and not only would I pay it happily, I'd be yelling "SHUT UP AND TAKE MY MONEY" as I bought additional credits just for that feature alone.

[teal raft]

You're right, but the fundemental reasons it doesn't work this way are complex and deep

and end up with "it can't work like that"

The way LLMs and Suno work there just...is infinite more complexity trying to get clean STEM tracks to train on but also to train them, somehow, to interact with each other

Might not be enough data to train, and if there is, might be an almost-impossible ask to somehow train it without it turning into noise

If it were easy, it would be done like this

would be better though, yeah

[hot trout]

I suppose that makes sense, now that I think about it more carefully. Out of curiosity, does anyone know of any reliable method of getting music transcribed? If I could make that happen, I could always effectively recreate my instrumentals in a DAW... not a perfect solution, but still one that would work.

[charred merlin]

Admittedly I know next to NOTHING about LANDR, so the above post was just what I'd picked up from industry conversation in the recent days, but I've been invited to check it out, so I'm going to try to use some of the unused Suno stuff from the past couple months that didn't end up being "polished", or that I abandoned for one reason or another. Tinda is my resident LANDR expert, she's taking a similar approach to it how I would have if I was able to drop the cash on the subscription on a whim -- hence the frequent recommendations for BandLab's free service and DistroKid (despite some of the stuff I guess I need to read more into)

I'm just glad to finally think I understand wtf blockchain and music could have to do with eachother 😅

[stiff sable]

yeah for sure! i'd be interested to hear your experiences with it actually and what its like! i do love LANDR's services and use a lot of them apart from the distro bcos they have rlly strict rules on what they accept and guidelines on your release data + cover art etc so im missing out on it purely for personal preference reasons at this point lol

[stiff sable]

this might be the most useless thing anyone could answer with, but the way i like to do it is use a stem separator first and bring them all into your DAW and just listen to each part, take notes on it and just recreate them one by one. identify the chords (you can use tools for this if you need to) and then writing the melodies, bassline etc by ear and then working on the sound design, layering etc.
imo its a very flexible and fluid way of working on it than having a strict transcription done first, but i might be misunderstanding what you're asking!

Woodhouse is in a mood 😭

[charred merlin]

@teal raft
I think part of the issue is a lot of people are becoming familiar with stems JUST at a point where we're now actually able to retrieve them with AI, so people who are used to hearing isolated tracks/stems resulting from having the original Master records, so they're wondering why Suno isn't able to listen to a song and pull a result with this level of detail(https://youtu.be/87FjkqtK67o and reverb out of a "dirty mixdown"

[hot trout]

I mean, I knew that audio was usually composed of different layers, it's how AI interacts with these things that I am still confused by. But yeah, in hindsight, there's no surprise there.

[charred merlin]

#📚┃suno-school message

I kinda tried to explain my understanding of it here -- from my interpretation, AI can be trained to take a picture > make it blurry > "understand" how to "remove blur properly".

After it figures that out, they can make up a random blurry picture that doesn't mean anything, and it might be good enough to "imagine" what the picture would be, even though the "picture" was just a random blur to begin with.

Just because Stable Diffusion learned how to take a picture of static and turn it into a picture, it still only handles it "from the surface", so while it can "imagine" a picture of a girl standing in front of the Eiffel Tower, it doesn't even bother to consider "what's behind the girl", because she's not actually in Paris taking a picture, it just looks that way from a surface level.

Similarly, Suno doesn't actually make songs by "imagining" each separate instrument track on top of one another, it just "imagines" the song all at once.

Suno being able to "look at" these songs and "imagine" what the different instruments would sound like on their own is A HUGE DEAL. AI music generation (beyond the existing algorithmic creation tools we've had for a while) in and of itself is literal science fiction magicand the currently ongoing revolution and the implications are now able to be so casually discussed by anyone that it's mind bogglingly different -- no wonder the establishment is utterly terrified.

So I 100% agree that it's really unsatisfying to separate the stems and be left with this "ugly" track, admittedly with which I have zero technical knowledge on how to properly "clean", beyond just trying to separate with a different tool, or waiting a month or two for the technology to jump another year's worth.

If you listen to the youtube link I shared just above, you can hear the slight vibrato in Marvin Gaye's voice, and the fadeout in the reverb -- little details that were completely lost in the original mastering process -- and it was only just a few years ago that they were able to find the original recordings to hear his vocals in isolation (before that, it was only ever experienced by the artist and the sound engineers)

Suffice to say, I'm not 100% satisfied with ANY service's splitting quality, but the fact that that technology exists, and is freely available without the need for a fancy computer rig? (https://www.bandlab.com/splitter)

2500% satisfied. Which is way over the recommended daily dosage of QoL

[hot trout]

I will check that out, thanks a ton!

[sturdy flax]

Interesting...

[viral lotus]

Sorry if you covered this already, but i'm hopping into the conversation midstream. I just got into the philippines a few hours ago!!

Anyway, does this stem separation allow us to do anything within suno that is useful, without using external tools?

@charred merlin Why is it that when i am on mobile, most of the links in showcase take me to the song on the suno website, where is on pc they are all embedded videos? Some of the links on mobile are still videos.

[charred merlin]

sorry, 1 sec my dude, sorting out last minute itinerary stuff while Mama Harber is in town

[alpine venture]

https://platform.openai.com/tokenizer

tool I've been using this run

I highly recommend it for getting a better feel for how the platform likely tokenizes

and then understand basically what tokens may have picked up meaning

when it comes to the composition. (it would help if we had a confirmation about what tokenizer of course)

[charred merlin]

WHOA lost track of time -- immediate thing that jumps out at me is being able to remove vocals from a track, and upload it into back into Suno to "Set the tone" for extended generations -- I know unless the track is 6-60s this doesn't work -- admittedly a hamfisted first suggestion, but I know a lot of people want to be able to keep uploads and their extensions public.

Less technical -- just stripping the vocals out of Suno songs you've made and turning them into karaoke for you to record your own vocals overtop -- don't stress to much about the fact that the instrumentals sound "weird" or "warbled" where the vocals have been "wish-boned" out, because your voice (or whatever AI voice you overlay) will cover that up for the most part. And soon enough the tech will have caught up where we'll feel silly if we groan too much about the immediate results of them squeezing a new feature out of the V3x series -- a less transparent company would have sat on the tool internally for months until it "passed the smell test" and by then there would be umpteen different competing services, but we're in the back seat to witness it all happening in real time, baby 😎

[charred merlin]

At least on iOS, I’m able to watch the embedded video or go to the site depending on where I click.

I don’t have an android to test if it’s the same there, though :/

[charred merlin]

dude -- I love seeing other people use tokenizers to try to "read suno's mind"
#📚┃suno-school message

this-is-why-i-love-hyphens-now

[alpine venture]

yea, I happened upon the hyphens by chance when messing with anti-

and slowly started using them more as time went on

but yea, now I'm onto finding isolated tokens

and trying to use them

to influence the sound and/or see if they have any meaning

like, if a-wave is something, I want to know

if that makes sense

I've already had some success

https://suno.com/song/b116841c-2a28-4f4a-98b9-d4c8ae1e456c

not pub

but it's a good example of how the model sees things

this takes the tokens that make up seminal and baroque, in certain cases (as the tokens change contextually) and combines them in a way that preserves that token breaking. Reliably produces dramatic scores, and it uses very little real estate.

[neat nexus]

So how does this work? You just dump your lyrics in and it what? Hyphenates stuff for best effect? Not being facetious, trying to understand what this all is.

[charred merlin]

the tokenizer?

[brittle apex]

I had thought you couldn't get the Udu sound in Suno, like this: #🗣️┃chill-chat message, but I accidentally did come across it in this extend: https://suno.com/song/6bb2120f-6270-49fd-954d-b5ed03515d5f

[charred merlin]

I shared a description in the tech-chat post linked in my comment, but essentially it shows how AI parses prompts

The same way that compilers turn code into binary, tokenizers show how our input is translated into "tokens" that the AI can understand

[neat nexus]

Yeah. Not quite sure the benefit or the process.

So you paste the lyrics as is?

Or the metatags too?

[charred merlin]

Can I do a quick screen share in DM? Way easier ❤️ I'd refer others to other times I'd described it, but I think I can sum it up in like 2 minutes

Totally cool if you're not able, #📚┃suno-school message this is a good starting point, then ask any followups 🤘

omg, I keep forgetting this is the future, and computers can do cool stuff, I'll just record a super short video

I swear to god I'd forget my head somewhere if it wasn't screwed on tight enough

[neat nexus]

I'd be willing to do a screenshare but due to my setup audio is not great in Discord.

[charred merlin]

No worries -- screenshare probably would have been technical overkill for what I wanted to share -- just sent a video, ask any followups here or in dm

Mama Harber and Mrs Harber are getting back in from a day trip together, so responses might be scattered when they arrive, but I'm happy to help

[sweet wigeon]

#💡┃suggestions-web #📚┃suno-school What if problem lies elsewhere. "Just because Stable Diffusion learned how to take a picture of static and turn it into a picture, it still only handles it "from the surface", so while it can "imagine" a picture of a girl standing in front of the Eiffel Tower, it doesn't even bother to consider "what's behind the girl",

" Bing's Copilot uses Designer API to create picture. If you pass your image, it just reads the image and describes it with its own words. It is not more exact than you would be if you would describe it by your own words directly in Designer. That's why I prefer to use the creator directly instead the stupid Coopilot, which is limited per 5 requests.

😍

So my idea is - what we really need to get better results and to prevent the wrong results we MUST have advanced settings. Suno can be inspired in Udio preferences - this is great thing.

😍

But more useful are the advanced settings. But that is still not enough. Because depending on WHO uses Suno, we need more types of advanced settings or options. As someone who wants to compose lyrics and songs I need some options which could even improve Sunos performance just by adding tools to create good lyrics, which would fit to preferred rhythm (shortening the time for production by texting by text-promp only, will save Suno's audio performance ). Once after my lyrics are great, then I can create my song with ease.So what I really need are some tools to create lyrics 😍 and/or rhythm 😍 so the text will fit the song to create good song and

😍 to have the advance settings which would include the following inputs:
... this to be continued in next post.

[oak kraken]

Copied from an email I sent to support, any help before hand is appreciated:
"Hello, I was just wondering if there was a way to add prompts to tell the AI what to do within the custom lyrics? I want to be able to switch from a male singing the song to a female singing the song with slightly different wording. Is there also a way to make it so that the singing ends once the lyrics are over so the AI doesn't repeat lines you didn't tell it to repeat? Thanks!"

[charred merlin]

Sorry to be a bother, would you be able to add line breaks between separate ideas…

…Like this? ❤️ 💪

I can tell that you put a lot of time and care into writing this to me, and I’d like to make sure that I’m able to better respond to each idea on its own ❤️

I might be a bit delayed to respond as I have family visiting tonight 😄

[amber vigil]

not always. Are you finding this happens on every gen of yours?

for me, it’s rare. I would like to see a link with original lyrics (ie non-edited) if you are willing.

[glacial scarab]

oh this is fantastic I've been experimenting with this as well, the mashing together in a way it still "gets it" but got sidetracked by other stuff

[amber vigil]

you did this style in v3 with no problems?

Does the issues happen all the time? Even if you regen?

Have you tried spamming [End]?

[brittle apex]

Interesting. Just messing around, but this is one of the better male vocals I've had: https://suno.com/song/0d38acc1-0525-4802-b000-56ae02e5b02f, and oh my the falsetto in this one https://suno.com/song/e414473c-6bad-42bd-8590-41bf2364fabd. This needs more exploration.

[viral lotus]

Problem is it is much, MUCH more difficult for a model to write music first and then fit lyrics to the rhythm then the other way around. Writing "good" lyrics in itself, without even considering music, is currently not possible in such an automated way. Chat gpt is absolutely horrible at writing lyrics unless you guide it substantially. It is not something where you can prompt a model in a few short sentences and have it spit out amazing lyrics. I have had some "fairly okay" lyrics spit out by chatgpt without much prompting, but this happens maybe one out of 50 or more generations, and even then the lyrics usually need to be modified substantially.

[mystic flume]

That's the biggest benefit of the premiere plan, being able to do hundreds of generations (or even thousands) gives you a much better result. As long as you have the patience to react to what it gives you and define your lyrics accordingly.

[mystic flume]

Yes, that approach works for everything. Bridges, coda, verses obviously. Refrains. Etc. I do it preemptively now so if I need to go back and regenerate or if I want to repeat a particular interpretation I have a much easier time.

[charred merlin]

imo, the inverse --lyrics can influence a song's style as well -- personal preference, but I create the songs with an instrumental intro to establish the song itself, then let that influence the extensions with the lyrics.

[mystic flume]

It's both, depending. Sometimes I start with lyrics. Other times I start with an instrumental. A few times I've had Suno generate nonsense lyrics and then wrote real ones to fit that template.

[faint garden]

Hello guys, I have a really important question, does someone know how I use the same generated id for make another music? Like I wanted to change some lyrics not only add, but same voice, same background music and etc, I appreciate any help

[mystic flume]

Use an "extend" — that's exactly what it's for.

Frequently I'll extend after the first few sections, once I get a structure I like, and use the same lyrics to make minor fixes to pronunciation or wording.

[modest oasis]

So with the new stems feature you separate the music and lyrics . Does that allow you to use the music you like with different lyrics ? Is it mixed inside Suno or do you take the 2 separate pieces and put them together in another program ?

[dusk kettle]

From what I can gather, you have to piece together the lyrics and instrumentals in a seperate program if you want to use your lyrics in another instrumental. Haven't seen anything showing you can merge an instrumental and lyric stem yet on the site. They might add it later on though.

[scarlet verge]

Hi guys, I'd like to ask you a question. Is there any way to better control the amount of time added to songs? Because I'd like to know in order to understand why certain songs are extended to 3 minutes or more and others that only have 59 seconds added.

[merry creek]

I generally try lyrics first. I try and put a great deal of thought into the rythm and syllable count of each sentence making sure they match. Rhyme scheme can help the flow. BUT some of the greatest things I've found is if I define what I'm hearing in my head musically (not notation, feel) Suno can get close.

One of my favorite songs that I need to work on for thousands of credits yet... I told Suno: Broaddway musical. ( It sets the instrument choice. ) followed by a rythm note (dissonate Staccato notes) setting style and finally (Swirling rage) for emotion. And I'm pretty happy with it. I then told that there were two voices older woman younger woman. And tried to give them names in the parameters.. that worked okay.

It obviously isn't what I would wish to write.. but I'm so out of practice in my piano and theory suno worked as a fine idea developing tool.

Mine own technical question is... does Suno regonize musical keys? F sharp D flat Major? What do I need to input (if it is a possibility) to get voices to sing at the same time two separate lyric lines in 3rds 5ths or whatever from each other?

If I can get the unison set down... voices wouldn't even be an issue... I hope.

[fallen zephyr]

Any best AI's to separate the Trackouts / Stems of song generated with Suno?

[sweet wigeon]

how to you make the feedlines on PC keyboard?

[sweet wigeon]

#💡┃suggestions-web #📚┃suno-school
(this is second part of my post about the advanced options)

😍 to have the advance settings which would include the following inputs:
🇹 ▶️ 1) tempo prompt,

🇹🇸 🎼 2) time signature prompt could be used for specifying of scale too,

🇷 🎶 3) rhythm (these prompt may be different for verses, bridge, chorus),

🇻 🎤 4) prompt for voices,

🇮 🎹 🪗 🎸🪈 5) prompt for instruments,

🇨 🇭 🇴🇷🇩🇸 6) Chords,

🇲 7) Melody - created with notes

🇱 🧩 8) Lyrics

🇸 9) Style tags

🇨 🌐 10) Country of origin like now I am interested with Latin-Cuba rhythms. These are the minimals.

[sweet wigeon]

#📚┃suno-school #💡┃suggestions-web I know guys, too many prompts sounds complicated, but really, these are basics if anyone who wants to create songs with AI, because we need to tell it exactly what to do otherwise the results are weird and unwanted. Like few days ago I randomly generated song like TOP 10 but disco and tuts tuts tuts syncopation... but that is not quite my style. I love things like classic, latin, Cuba, flamenco and experimenting with India Rock and Baroko (Bach) music. Every singer is using some singing technique which should be specified - well described - in the vocals section. Need to say where to use that technique... And many more things. I learned how to prevent to make Suno using the reverb effect or metal noise stuff like cymbals or REMO spring drum... Just by specifying the singing technique. The thing is - to make good song, need to learn how to do music, the terms around it, the theory what scales to use to play some melody, etc...

Still I have forgot the need

🤺 to specify the explicity of mood interpratation (the expression). on/off button. Normally the AI changes the expression (alias interpretation) based on the lyrics words. So when you have a list of words like I did ... a training lesson of Ukrainian language. Something like doctors/medics, disease, child death, divorce, the AI will change the expression of the mood and inserts 😫 dark 😫 or 🥶 melancholic 🥶 feelings. And you cannot turn this off. For songs like this when I want to create song about learning the vocabulary or basic language phrases I just need to turn the expression off. Otherewise the result is very dark, sad, stupid song full of metal noice or reverb. Which I really don't want. So when the mood explicit interpretation would be off, then Suno should read from the prompt what mood to create for all the song. Like happy or G Dur, F Dur, using viola and flute could generate very happy song even with the "bad" words.

[graceful condor]

i have used fadr online for some good results

[past aurora]

Hi guys, does anyone have any tip on getting better dry clean vocals without the autotune-like sound?

[crimson terrace]

Avoid trap/rap/hiphop
And try
Bedroom cover
Youtube cover
Acoustic cover

Basically anything known that is not produced in a studio

[past aurora]

@crimson terrace Thank you. I will these a try 👍

[tropic fox]

How do you get it to generate nature sounds like rain or thunder?

[charred merlin]

Thank you for editing your post ❤️ I got my coffee ready, I've managed to convince the cat to sleep anywhere BUT my keyboard, and I'm listening to some Chick Corea -- time to start the day ❤️

I agree with all of this! You raise a very good point that WHO is using Suno matters very much. Some people only ever use "Simple Mode", because being asked to choose their own lyrics or "styles" might be something they've never had to think about before 🤔 To some people, metal, hardcore, punk, indie would all be "rock", and while it isn't very descriptive, thy aren't completely wrong.

It must be very hard to design a product that everyone can use in the same way -- whether they want very simple tool, or very advanced tools (like you and me 💪) -- too many advanced tools, and new users might get confused and think "they can't understand", but not enough advanced tools, and you and me feel like "it's only a toy".

I like how you make a comparison between Suno and Bing Copilot images! 💡 When I want to make a "quick AI image", I use the Copilot app because I believe what I want to do is easier on the app -- it is very helpful having different choices on how to use Image Creator 🔍 . With Suno, the app and the website can both create music, but the app is focused on new-and-casual users -- it seems like the suno.com website is where they are focusing on having "the most tools".

[charred merlin]

Twisted perceptions gave a great description -- there isn't a means of chopping specific sections of a song off directly inside Suno as of yet (but #🗳┃mini-polls message means they're asking how badly people want it...).

However, if you remove the vocals and upload 6-60s of the instrumental into the audio upload and extend off of it, it will use that context and most likely give you a very similar song, albeit with new lyrics.

[charred merlin]

@neat nexus woodhouse didn't want me posting this, and was just saying "no no no we don't like that", so I can't even parse his "error message" to figure out what I'm supposed change. Maybe it's a rule against the code blocks? Either way, I just posted the picture, but please delete it if it goes against posting rules or something 😦

[neat nexus]

It probably was the code. I'll mention it in the guide section but you're probably fine.

[charred merlin]

Not yet, at least not as a "command" in that sense -- it reads it as a suggestion, which unfortunately might not get the expected results unless the "suggestion" makes Suno think of a song that matches the criteria.

Not to ramble on a tangentially related point, but kinda neat that LANDR is trying to make this kinda stuff possible with the Fair Trade AI thing -- they want to include this sort of metadata in their song library's training data so that when you ask for "120 bpm, concert b flat, female vocals and flute duet", it could specifically reference songs that have those metadata readily present.

[charred merlin]

I'm used to working around Woodhouse, being able to "mask" links just helped with readability, but that's not a huge issue. I don't use actual code blocks very often, usually just code snippets, but just as a one-off for the sake of making a table it helped for readability

[neat nexus]

Right. The actual message wass repeating characters. So it looks like it was actually the | that it didn't like.

[charred merlin]

Ahhh, that makes more sense. Please apologize to Woodhouse for any aspersions on my part -- he means well, and makes one hell of a daquiri. I can't stay angry at him for long ❤️

[charred merlin]

100% emphatic recommendation for BandLab -- competitively powerful, free web-based 4-track splitter, DAW, AI mastering -- all with zero limits on usage, all available with a free account (no payment details or obligation unless you decide to check out their paid membership, which you'll NEVER be forced to use) #📚┃suno-school message

[rare oak]

Yoo, do you guys have a tool to remove background noises from your tracks ?

[charred merlin]

@unique stone @naive quail, I could have sworn one of you had sent me a song with weather effects in the intro, or shared it somewhere, but I couldn't try it down. Any idea what I'm trying to remember?

[charred merlin]

What types of background noises? hiss / pops / clicks / cats meowing?

It's not present in Suno, but there's lot of noise removal tools built into commonly used DAWs as plugins, etc

[charred merlin]

If you hold the shift key when you press enter, it allows you to start a new line ❤️

💪

[rare oak]

like a continuous hiss when there is a lot of elements in the track

[charred merlin]

I'd suggest looking into low-filter passes in a DAW that you're comfortable with, even a quick youtube tutorial can help teach the basic-basics of it to dip your toes in the water, so at least you know what methods might be overkill for what you're trying to accomplish

I'm recently learning JUST how oldschool I am for recommending Audacity still, so I might wait until someone who actually knows what they're talking about can recommend something like cakewalk or something

😅

[rare oak]

Perfect thanks for your help

Audacity is the one I'm using lmao

Currently learning how to use Ableton

[charred merlin]

You'll see me mention BandLab a lot because I'm really fond of their free web-suite, and I just learned that CakeWalk is made by the same team, plus I think JawKnee or Jon Fly use it as well, so I wanted to look into it.

But lots of cool stuff happening LANDR wise, so I'm hoping to get a lot more familiar with that in the near future

[charred merlin]

Any first impressions coming from Audacity? I like how Audacity isn't flashy, feels intuitive and laid out as I would expect -- never spending more than a couple seconds trying to "guess" where a feature I need should be, as I'm not such a frequent user that all of the best-avenues to use a method are second nature

[rare oak]

Well I like to use Audacity, it's just SIMPLE to use, and not time-consuming when you want to do just a little edition

Surely it's limited at some point, but for the moment it's all I need

[naive quail]

I haven't had much luck with v3.5, but I have had some with v3, at least at the beginning of a track... and then extending from that and changing the styles in the extensions: https://suno.com/song/31a91d5a-1222-4a13-a1fb-6c3df8ca7ccd

[naive quail]

I keep a number of my little experiments like that in a playlist: https://suno.com/playlist/8fc9cb0e-1a70-40e7-92b7-0f762dd2c109

...tho most get deleted so that I can keep my Library clean.

[hasty stag]

Can relate 🥲

[sinful merlin]

I had this idea for a wholesome song about a girl growing up while her nerd father shows her games and the internet from dialup to tiktok but idk how to make suno make the sound effects, i would have to make them manually correct?

or is there a trick I don't know about

i want it to make the dial up sound

[obsidian skiff]

What does this even mean?

[rare oak]

did you already uploaded it before ?

or maybe it's anti-plagiat

[obsidian skiff]

That is the first time I used that file

[hollow hare]

It means that parts of your song got a hit against ContentID or whatever software Suno uses to check for copyright material.

It is actually because of said system that a few people found out that someone stole their song from Suno (or was linked here, etc.) and distributed it

[prisma panther]

Hi there, i would like to know if it's possible to download the "multiple audio Tracks" from the created song or is it a unique stereo track. And second, what is the audio format ? Wav, Aiff, Mp3 and the rate : 44.1khz, 96khz ? Would be very happy to know.

[stiff sable]

you could potentially add tags like [Phone dial-up tone SFX effect] in the lyrics box at the point you want the sound, but these are just mere suggestions to Suno and the chance of you getting what you want (or at all) is extremely slim, but I've seen it work form some ppl so it could be worth a try.

[spare stump]

greetings ...quick question for the geniuses

[rare oak]

SOmeone knows a simple tool to enhance the bass of a track ? Not enought 400hz on mine :c

[leaden lark]

What’s a style prompt for female vocals in a genre like dream pop?

Like I can get good results already, but something like this https://suno.com/song/0cc062ea-136f-4522-9622-07b9f627406b

[charred merlin]

I talked about it elsewhere in this thread, if you search for "Fair Trade AI", where a company is trying to make ALL of this possible -- I have complete faith that Suno will only continue to get better and better, but it is very exciting to think that there are very smart and influential people currently trying to solve the exact problems that you bring up! 🙌

Keep in mind that things are moving SO fast it's hard to know what to expect from month-to-month 😄 Technology and generative-AI are at such an exciting place that it's no longer a matter of trying to figure out HOW to solve problems, but which of all of the possible ways to solve a problem will help you continue to make things better -- it's not about "fixing cracks" in a product anymore, it's trying to figure out how the NEXT product can be 5x better than the one before it -- and now we don't need to wait years between big changes, @worthy grotto are a really cool team, and they seem like they really care about making sure they make what the people want

💙

[charred merlin]

The song has been flagged as being owned by someone else -- if you own 100% and exclusive rights to the clip you're uploading, you might need to contact [e-mail] to clarify the matter (because false-positives can happen! They want to help you resolve the issue, but you need to work with them ❤️ )

If you DON'T own the clip, that error is just to try to prevent you from possibly getting your song or account taken down by a copyright claim (and since things are changing EVERY DAY in the AI-legal world, possibly to prevent you from being taken to court?)

@vast monolith if you waited for me to stop talking, you'd never have a chance to join the conversation 😛

[leaden lark]

I’ll have to contact them because I think I got something similar and the audio was generated from Suno

Any help on this? I can’t make it public https://suno.com/song/d62d1d32-c482-4c7b-b6e3-ceaf28737c1f

[hollow hare]

If you used Audio Upload, Suno tries to detect if there was any vocal in the clip.

If there was—or from a false positive—Suno disables the ability of the song to go Public on the website in case there may be a privacy violation (e.g. no consent from a third party present in the clip).

[leaden lark]

Thanks for that, so should I contact support? The song did have vocals that I split in FL studio before using that audio as the audio upload base

[hollow hare]

If the clip you uploaded has no vocals, the team may be interested to help reduce the false positive.

Just in case: if the audio clip was not produced by you or through AI (e.g. taken from someone else's content that you had no rights over), e.g. a copyright violation, you may be penalized which can include deletion of the song if not the termination of your account.

[modest juniper]

I have to say that I really like Suno compared to Udio. Udio is a pain in the @ss. Yeah they have inpainting that I thought might work well. It does not at all. I use my own recordings that I have made over the years and I would like to change bits of my songs such as guitar solo's and leave the rest. My solo's are much to be desired. If Suno could add inpainting and the ability to load the whole song so that one could change just the parts of the song would be fantastic for me.

[charred merlin]

Ah dude -- I only recently learned that UX Research and Scrum Mastery were "things", so whether or not your response was directed toward me, I'm gonna glom onto it. 😅

I've thought a lot about the different personas regarding Suno, from the people who are embarrassed to hear themselves sing in the shower, to people like myself who are willing to zoom into tracks at 2500% view to manual move dot-samples in the wave form, to users who know EXACTLY what they want with hyperspecificity because anything else doesn't fit into their preferred workflow. NONE of them are using the service incorrectly, and as long as they're able to be meaningfully expressed, none of their opinions should be discounted -- so how do you "optimize" the platform for each of them, ideally without making sacrifices for any of their experiences? New, casual, superusers, and stakeholders all have entirely different expectations of how the service should currently work, and what the focus moving forward should be.

The past few weeks have been really been eye-opening -- getting to know so many Sunovators on a personal level because they're excited to talk about what Suno means to them, or how being able to enter the "creative space" through the platform has opened doors and introduced them to new hobbies or industries they never would have experienced before.

Quick related anecdote -- I played guitar and sang in a band for a good number of years, and while we recorded stuff just for ourselves and our friends, I never would have dreamed that our music could have been played on Spotify or Apple Music, simply because I had convinced myself "that was for real musicians" -- even just personally speaking, Suno has done a LOT for me, and I'm not the only one.

[charred merlin]

God, you're a Mensch. 🍻

btw, you should rename that crazy instrumental mixup "happy accident" in honour of Bob Ross

[charred merlin]

The page doesn't seem to be active on the community wiki anymore, but I grabbed a screenshot from web.archive.org, if can still help provide guidance for the moment ❤️

It seems like the article is written from the perspective of instances where your songs generated from Suno are flagged by other services, but I suppose it should still provide some sort of guidance even in the instance that Suno is both creating and flagging it 😅

[hollow hare]

Definitely sounds like a response to people who use TuneCore or uploaded to SoundCloud because there were several people asking about that (TuneCore was leaning more towards having licenses iirc).

I was trying to find past convos I had seen about those situations but only found these two cases in my brief search:

• SoundCloud: #🗣️┃chill-chat message
• TuneCore: #💬┃general-chat message

[leaden lark]

Thanks for this, I’ll submit a review when I get home from work as everything about the song was using Suno

[leaden lark]

Everything was 100% from Suno and the audio base was the song I chopped up in FL Studio for the base

[tropic fox]

How many attempts did it take to get the thunder?

Would be nice to be able to get it consistently

[hollow hare]

Great to know but, for clarification, I was not accusing that (as the copyright filter would have caught it if that was the case).

I think someone once said a didgeridoo had triggered the vocal detection. The system still needs fine-tuning so submitting your clip can help the team with that if it had no vocal-like sound

[leaden lark]

Okay, so in the ticket, link to the song and briefly describe what the issue is?

[hollow hare]

Er, what? I did not get what you are trying to ask

[leaden lark]

I was just asking when I submit my ticket, should I link the song in it?

[tropic fox]

There is a bug in suno, if you extend and you enter the same style info then will confuse itself and complain it's copy write even though it's your own stuff from the previous input

[jaunty coral]

Is anyone else having an issue where the 2nd chorus is nothing like the 1st chorus? I've been building this song by "extending" it and now that I've gotten a couple of verses finished and want to swing back to the chorus, the singer is singing it completely differently and I haven't been able to get it to match

[slender tide]

yeah that can happen as the AI is kind of making it up as it goes based off of about the last minute of audio from the extended spot. so it doesnt exactly know what it did for the last chorus

[hollow hare]

Ah, it would not hurt to do so.

As I mentioned, the system detected a vocal-like sound which disables the Public option.

Providing evidence that the clip came from a portion of another Suno-generated clip (with the time frame of when the clipped sound was taken) will help the investigation team disable that restriction if they have the ability to do so.

[slender tide]

you could always drop it into a DAW and do a little copy/paste

[jaunty coral]

In that case, do you think I could "Get full song" and extend from there?

[charred merlin]

@leaden lark , I think there was a crossover, my reply was to another user who had the error "your content has been flagged as an existing work of art"

[leaden lark]

Thank for the help! I’ll go back into FL studio when I get home from work to figure exactly where I did the cuts for the audio base

[slender tide]

that's usually how I do between extends. not sure if it helps. If I definitely have a perferred chorus (or section) it did somewhere else, i will download the audio and "finish" the track in DAW by cutting it up and blending it in myself. so for your part, cutting out the chorus you like and then blending it into the second chorus spot

[frigid pendant]

вовоо ¯_(ツ)_/¯

[jaunty coral]

I'll try it and see how it goes. Thanks!

[leaden lark]

My bad, at least it seems I can get this resolved once I get home

[charred merlin]

I apologize for any confusion as a part of my response 😅 Shouldn't need to reach out to support for that, as false positives can occur. For instance, digeridoo can reasonably "diffuse" into "droning vocals".

I've heard of people being able to make minor adjustments to their owned-content and resubmitting to bypass the vocal flag

[slender tide]

if you have some background in DAW work - just set the BPM of the project to fit the song's speed (I use FL Studio) and it should be pretty easy to cut on beat and slide it in. and it should blend through pretty seamlessly

[charred merlin]

Sorry for the delay, Mama Harber's last couple nights in town before she flies back out, so spending as much time as I can with her ❤️ (sorry - not sorry kinda thing, but I'll respond to everyone who the phenomenal team of guides hasn't slam dunked ❤️

[hollow hare]

Do not be sorry, that is much more important than being here 🫂

[leaden lark]

It’s alright! I can wait for the manual review

[loud kernel]

Any idea when we might get a audio fidelity upgrade? 👀

[jaunty coral]

I haven't the slightest experience with any of that, but there's YouTube! lol

[vocal meteor]

Does Suno have an API? Or am I limited to the web form?

[slender tide]

web and iOS app currently

[naive quail]

I was able to get it fairly consistently at the time with that same style and with [Intro] and [Ambient Noise Interlude] on the first two lines. The intro line probably wasn't necessary

[tropic fox]

Just tried 14 times. No dice. lol

[naive quail]

just did a couple rolls... this is the first one. A bit rainy... less thunder. Other effects. https://suno.com/song/09d7ecf1-2153-4378-90f0-e059a6bb6bcd

[sweet wigeon]

Thank you for reading my suggestion list and my reply. Few notes from me.

⏩ too many advanced tools, and new users might get confused and think "they can't understand", but not enough advanced tools, and you and me feel like "it's only a toy". ⏪

There might be option in Preferences where they can turn this prompts. So the newbie would have three prompts, but the expert will turn all the prompts on.

Newbie might think it's easy to create song good song. While this is his/her perspective of view and doesn't see that it is good for him only, not for sale or even radio broadcast.

It's like if I would create song with title "Baroko concert with piano and violin"
with description "inspired by Bach", and I would sell it 6$ dollars for album, but then comes some pro musician who knows classic, and will complain, this is not inspired by Bach, there is loads of music from Mozzart.

I wonder if Suno can think about that the app can be used by passionate beginners who want to start playing instruments or singing ... to learn to play or to start to play because it can motivate people to start to learn some theory, and then to do it in practice.

[remote flume]

You guys/gals know why suno is not singing part of my song, but ibstead starting with the last half of my song?

For example:
[verse]
I love you, don't you know,
Oh it's true, my eyes they show,
Could be us in the snow, down low,
So, lalala let us glow, in the dark

the ai is skipping the first two lines and only singing the last two no matter how much time i prompt it?

[spark valley]

People, my application on the server is not working! What to do?

First there is an interaction error and then the application is not available (((

[unique stone]

If you mean the /chirp bot, that has been retired a few months ago and has been replaced by the website at https://suno.com

[neat nexus]

Just saw this, sorry for the delayed response.

Try getting rid of the comas at the end of the lines. You can leave the ones in the middle of a line.

If that doesn't help, test seeing what happens if you space out the lines. Like:

`[verse]
I love you, don't you know

Oh it's true, my eyes they show

Could be us in the snow, down low

So, lalala let us glow, in the dark`

Are you continuing from another part or starting with this part? If extending try changing the time you are extending from to see if that makes a difference.

See if adding vocalizations helps before the start of the verse metatag. See if a spacer makes a difference. I use (...) for a spacer as needed. YMMV so keep trying stuff.

[modest coral]

Are Suggestive songs aloud on this discord? Like.
I don't mean like XXX lol
But I wrote a song inspired off one of my stories... My story is kind of a dark fantasy like game of thrones or the house of the dragon.. It explores explicit themes.. But the song itself, I would say is p.G thirteen

https://www.instagram.com/reel/C94Q8_Axu_a/?igsh=MXgwcDFjYzZpcnJ6MA==

I posted it here if you guys want to look at and listen to the lyrics and see if it's appropriate?
If it's not I will not post it

queenofobscurepairings

PLEASE FLIP SIDEWAYS TO VIEW
Art by @moe_machine
Original song written by myself!
From Pestilence's point of view

Instagram

And I will also delete it from this chat too once you guys are done looking at it the link I mean

[spark valley]

Thanks. But others somehow post their tracks here through this bot?! Or not?

[spark valley]

And by the way, what is the point then in this server and bot, which was closed but not removed?

[sinful merlin]

anyone knows how to do reliable fx? I want to make a song with the lore songs of the internet

[unique stone]

The songs that get posted here are generated on the website. Users often paste the links to their Suno songs here and discuss making songs, report problems, suggest features etc.

[spark valley]

I agree with you. Sorry, didn't think about it. But it's so INCONVENIENT. I read about Discord on the Internet, I thought it would be more convenient than on the website. Since the application is only on iOS, then at least this is an option. Sorry I'm late. Well then, let's poke around on the sites like in the good old days.

[unique stone]

Assuming this is a Suno song, a link to the song on the Suno website be preferable to be able to easily check the lyrics.

[modest coral]

Its not on the site anymore. I made my own custom lyric video

[spark valley]

• I read that there seem to be several servers with Suno, but I never found it. They always throw it away just for this one. Apparently the last one remained or was combined into one, maybe.

[unique stone]

There may have been some unofficial ones. This is the official one.

[spark valley]

Thanks for answering ☺️

[sweet wigeon]

Your app is not on their server, this is client communicates with their server. What's your browser version?

[sweet wigeon]

You'd need to show complete lyrics to tell. Show more verses.

[sweet wigeon]

Just convert the frequencies of falling rain drops into the notes. It's easy, but I'm not gonna share my best results.

[lucid valve]

@charred merlin heyy, thank you so much for the advice about contacting routenote to edit my release, today i received this and it looks like now everythings gonna be fine to release my album hahah, tysm 🫶

[spark valley]

Chrome 127

[mint flare]

hello all - sorry for the newb question, but I can't figure out why I am not getting any audio from my suno - like the same browser is happily playing youtube if I want, but there is zero audio coming from suno - even on tracks that I already made that worked great. ideas? I've checked it with soundsource and there's genuinely nothing being played.

[amber vigil]

@charred merlin is the man!

[alpine venture]

https://suno.com/song/87d41673-f39f-430c-a2dd-5ee0aa53af78

thought some people might find this output interesting

I've been working some new prompt elements

w/my new 10k

[teal raft]

prompt me ein papa

[alpine venture]

it's very very sparse

it's not good btw

but it's nearly all drum

and just some guy

Jonathan fly style

but without field

https://suno.com/song/9e92a99b-0cca-4b36-b38c-36dc002327eb

and the transfer to some puuuuunk

best woman's voice

I've ever gotten I think

https://suno.com/song/6dbc634a-d60f-4939-8713-09a9af1ed39c

yaaaaaaall this one is straaaaaange

[raven turtle]

When I get this a good old reboot of your device usually does the trick.

[sweet wigeon]

if your tab is not muted, your speakers work, then report the files to Suno as Bug.

[teal raft]

dang, so, yolkhead

I swiped your style prompt and slammed some lyrics I have together and uhh

p. neat

[alpine venture]

yea, I'll def be saving it for continuation especially

it's good if I need something weird

[leaden lark]

I tried to remake one of my first songs, thoughts or feedback? https://suno.com/song/f489a23a-92f1-4266-9bee-197b2c3d3d0b

[golden abyss]

Allright, weird one. Ghost in the Suno?? my roommate was working yesterday, and a song appeared in his library, he did not make it, neither did i. now the thing is, where the heck did it came from??!?!?

[lavish forge]

this is an italian song made with Suno

[hollow hare]

Hello, I understand you are excited to share your music but it would be more appropriate to link your music at #🐣┃suno-showcase, unless you have a technical discussion question about the piece?

In case you were unaware, mobile Discord does not play MP3s (and downloading unknown files can be risky with potential exploits which has happened in the past with mp3s).

Thus, as I mentioned earlier, it is best to share the link for the creation (or upload the video but only posting the link will allow you to see how many people listened to your music).

[earnest ledge]

Is anyone else having issues where the song doesn’t fully complete after singing a tune and then having Suno generate the song?

[hollow hare]

Hi, v3.5 is well known to be unable to end songs, if that is what you meant? To forcefully stop the music (if you are intending to make it public on Suno), you can do the following

How do I end a song in Suno 3.5?

Extend where you want to end with the new extension set to either v2.0 or v3.0 when using ending tags (e.g. [end] in lyric box and "end, short" as the style tag) as they are much more reliable for ending than 3.5.

(Short to help end it quickly in hope for credit refund)

If you are not intending to share publicly on Suno, you can use DAWs like Audacity, or even several web-based DAWs, to easily trim off the excess portions

[hollow hare]

Iirc, at times, suno might accidentally load the trending (or was it new songs) list. It is a simple glitch and refreshing the browser will fix it 🙂

[bright coral]

🤓 Are there any tips/methods to get songs when extended to use new lyrics added? When I extend sometimes the lyrics it will reference are prior ones. I have attempted to play with where the new words are pasted in the order of the existing lyrics but it seems to get the old words 😅 🤔

[lavish forge]

Ok sorry

[golden abyss]

actual song, with lyrics, under his account name, (still the default name tho)

[glossy zephyr]

Hard to control really, but as long as you remove the previous lyric sections, you have a better chance that way

[lone oak]

Why do I feel that there is something wrong with the Style of Music of 3.5V? The chances of the style and voice being wrong are too high. The male voices generated today are all female voices. I am really frustrated.

[proud compass]

Can someone explain the plays counter. I noticed one of my tracks has been sitting at 50 plays all day. I know ive listened to it at least 4 times. So at the bare minimum it should be showing 54.

[opaque tree]

Greetings. I am aware this question was probably asked and answered many times: Can Suno regenerate the song with changed text? I am very happy with one of my creation but i would like to change just two words in my lyrics so i can be completely satisfied with it, thank you for your time.

[alpine venture]

does anyone have a definite insight in what tokenizer is used?

I think I "maybe" have a p big new and useful observation in how the model views prompts

buuuut, not enough data yet

so it could be coincidence

but it does seem that the model has a tendency to weight evenly across tokens, but the fun thing is this means if you arrange your prompt so that something is more tokens than usual, this can increase it's relative weight within how it is evaluated

you can use random characters to manipulate this

to alter how the tokenizer perceives the prompt

likely, this is why some people, effectively acting on muscle memory or sort of "gut" model comprehension, were having results with certain character combinations

it also does seem to have some impacts on generalization ie overall composutional quality

so that's interesting

[teal raft]

they specifically do not mention anything that goes that direction with how suno works

I've asked 😏

[alpine venture]

it's gotta be openai

everyone uses it

I mean, like, not everyone

but mostly

[teal raft]

Guessing the 'interpeter' is running through openAI yeah

Considering the 'write me lyrics pls' does

[alpine venture]

yea, but also even if bot, if I'm not mistaken their tokenizer is open

I honestly haven't looked too much into this component

[teal raft]

tony once said esoterically "bark is open source~~~~~~~~~~~~~~~~~~~~~~~~~~"

[alpine venture]

though, so I'm out of my depth

[teal raft]

same

[alpine venture]

I think I'm right

this would explain why people believe they can weight by using bracket

it adds tokens in certain configurations

I suspect that people can also use this to get more clarity at the expense of control

just load all white space with diverse special characters

https://suno.com/song/9d3ea71b-acc9-4520-9458-35ba2797fbae

I broke him again

[teal raft]

So you mean leave no empty space

[alpine venture]

https://suno.com/song/4d1521a9-720f-4c4e-81d9-92d23c8008fc

this is a more coherent version of the above insanity

also image gen has gotten good

on suni

suno

https://suno.com/song/88fb619f-a26d-47ff-a2e5-bee4fd10f563

well well well

what have we here

a squeaky boy

Basically I think it's worth testing minimizing tokens vs maximizing

and see what happens

[lone oak]

Is there something wrong with the AI ​​now? The prompt word is a male voice, but dozens of female voices were generated.

[earnest ledge]

I’m using the app at the moment and it happens when I use the microphone to record my voice to sing, the AI doesn’t generate anything.

[late meadow]

I'm trying to add to my GPT instructions. I use this cause i frequently have to delete sessions after 3 songs. if anyone has the time to read and add to them please to so because i want to make good music please ❤️ this is is the most resent test and i made public this time. https://suno.com/song/07253889-f8aa-44c6-a3d9-0dceb480cc3c suno ai doesn't get it 100% but it usually comes close

[viral moss]

How well does using regular " ( ) " brackets work? In what instance would you use them?

I've successfully put one sound effect in a sound using them, but most of the time they're spoken out loud or ignored

[late meadow]

hmm idk when i tested it's prioritized like {Primary}Subsidary there is no actual method to this

i've tried coding the styles it works but it's really unneccesary

yeah for sound effects use [] square brackets suno will less likely speak those. the important thing is consistency , because there is no actuall format if you do everything the same through out the ai will intpret it as such

[viral moss]

I've never gotten the [] brackets to work in any similar instance :/ always ignored

[viral moss]

Very true on the consistency bit, I've noticed.
It's just such a wide range of options of how it'll interpret it.

[late meadow]

yeah you have to use musical elements. if it's anything suno doesn't recognize or easly interpret as a musical element it will get ignored.. or worse spoken. and what i mean about consistency is that if you do like [guitar strings, pluck] you can't use square brackets for anything but musical instruments throughout the lyrics

[viral moss]

Adding [spoken word] and then writing out the "scene" seems to work a bit, I've basically cut off a song due to the singer being interrupted mid singing

[late meadow]

oh yup XD

[viral moss]

And one time I got someone to vomit during the bridge

That took a lot of tries

[late meadow]

whoah! share vomit lol

[viral moss]

https://suno.com/song/ffefb914-4aa3-40bb-bdca-1057517bebc5 around the 2 minute mark, followed by an "oh god...."

[late meadow]

lol nice XD

[viral moss]

I was so surprised it worked

Also got a weird 20 seconds of silence at the end there. All one gen though, didn't need to extend it

[sweet wigeon]

I often use Liama 3 for composing - it's good. Very similar to ChatGPT except if you want to communicate in different language than English you need explicitly specify that. I can tell you a lot about rhythms in popular music if you want some good song. I asked for some specific song with 7/4 rhythms and ChatGPT guessed 8/4 or 4/4, LIama was much better (closer).

[alpine venture]

in this case I'm referring to the "composer" which presumably creates the melody; for lyrics though I agree, LLama is more creative, I used a process of ping ponging and unnatural selection to write the lyrics for this one:
https://suno.com/song/677e3f18-bcd8-41ae-a458-fa25faf6ba6f

I'm sick atm, I was planning on maybe streaming tonight and trying to sort of "officialize" a bot and method

but my brain is at about 30%

ideal is closer to 70 for creative endeavors, right now im just dumb

realistically though, yea, Idk why so many people are scared to use ai for writing, it seems people take it as a "mark of pride" that they write their own lyrics, when were literally using ai to generate music

[sweet wigeon]

I mean compose as creating scales and chords and notes, melodies and such instructions for "lyrics".

[alpine venture]

like, if tools will help me compose better poems, ima use em

[alpine venture]

ah v cool

does is interpret them?? Plz, share, this is something I didn't know

like, do you tend to get a good response from your output?

[alpine venture]

haha, your profile is Rashi

[sweet wigeon]

You can ask LLama for specific melodies, you just must write what you need. I have created good music for rain drops in specific scales expressed with classical music (I asked for violin and similar things). You just need to learn the theory and terms to understand how to communicate with LLM.

[alpine venture]

he's v popular

[sweet wigeon]

Yet it might happen that the melody can be in lower octave than you wish, so you just modify the notes or the octave, or the scale.

[alpine venture]

sorry, I meant rashi, he's a famous 18th century rabbi, likely the most famous

[teal raft]

I've never heard of him
is not jewish

[alpine venture]

just a random fact from your profile Pic haha; anyway, thanks for the info

[alpine venture]

yea, he's a bit niche haha

[teal raft]

I'll tell you all about Toyotas, ok

[alpine venture]

growing up I thought I was the funniest kid on earth bc I'd always respond "Toy Yoda?" like it was some profound insight

[teal raft]

[alpine venture]

as it turns out, I was just a dad all along

[teal raft]

As an idiot kid, that was your right

Now as a dad, it is also your right

[sweet wigeon]

I love what I got. But as a free user I got just 5 tries per day. So if the music is short like 2 minutes, and then i extend 3 times, I got 5 minutes of very beautiful music without voices, but divided into 3 portions.

[silent viper]

Does Suno follow all this information?
So far I haven't been able to, and I haven't seen anyone who can train the platform to follow so many details and generate the music as the user establishes in the "lyrics" and "style of music" fields.

If you managed to do so, please provide the link to the generated music.

So far I have not been able to determine the chord progression or the BPM.

If you have managed to do so, please share the link to the generated music.

I would really like to learn from your experience.

[tight spade]

@silent viper Look at my lyrics in top of song , usally helps with consistency. https://suno.com/song/b6659bd2-c6a2-4010-8922-a532515cabe9

still trying to nail cords tho

[alpine venture]

https://suno.com/song/af411b2c-caf1-4e6a-8233-49621dd12708

[silent viper]

Yes.
I've been following your profile for some time.
But I see that there is a lot of information and most of it has no effect:
"orchestral", "violin", "brass", "atmospheric,"

[alpine venture]

getting interesting output here

https://suno.com/song/1c0a76bc-1f36-4f42-966a-7e5ca25c7bfb

holy cow

major beastie boy

but like, Bubblegum

beastie gum

[hearty lotus]

Guys where is the key

[alpine venture]

these progressions are so much closer to modest mouse

[tight spade]

hitand miss all depends on the genera

[alpine venture]

than anything I've ever done naturally

without extensions

[viral lotus]

Can anyone tell me... I am trying to mass remove songs from my playlist. I can select them all with ctrl-a, but whenever i attempt to remove them from the playlist it selects only the one that i click to remove from playlist. How do i remove them all at once?

[teal raft]

What's your logic for building these prompts

[alpine venture]

I went into exhaustive detail with P on this, but basically it has to do with overfitting

and smoothing it out

simplest way is think of the model as an autistic savant with severe adhd

and were trying to bring its attention

to the right arra

area

[silent viper]

I love this tags on style of music's camp

[alpine venture]

but the thing is, she/he sees the world in a way that is totally alien to us, so our best bet is to exhaustive explain what we want

much like how we teach irl

where I might reach my kids about numbers three different ways before it clicks

so here, we want a lot of tokens that basically attack the same "area" but from different angles

and then the idea with the special characters

has go do with, possibly, the way the model tokenizes

but also with the core idea of getting a broader array of tokens

overall

which in turn basicslly improves the model's attnetion

I didn't do this before bc I wasn't sure the composer was an llm

and thus weird tokens might ne literally meaningless

but I've become convinced it is

based on its behavior

in how it interprets lyrics

[tight spade]

@silent viper im still learning the prompts i find all my prompts from , sunoaiwiki.com

suno.wiki

[alpine venture]

oh, but now I'm using the okenizer too

so ungun

for example, is related to how the model sees nigun

I'm tokens

in tokens

so it's all a part of testing parts of words

thus is assuming it tokenizes similarly to openai, or thexsame as

[tight spade]

@ivory pawn https://suno.com/song/3811b2c9-d442-44bd-ae45-9ce1556321ae , check promps i was able to orchestra piano violin working

missed spelled and forgot the "]" and also put words in wrong spot.

KISS. Keep, It , simple , stupid... i have hard time doing this

[viral lotus]

Sorry, i'm coming in mid conversation. Does encoding like this work on a regular basis? Have you tested the various parameters?

[tight spade]

@viral lotus yes

@viral lotus check this

brb

@viral lotus it takes some time working them out , but most prompts work.

iv used them in diffrent syles of music sucsefully ill giv e you guys a contrast of what things sould like vs diffrent prompts if that will help

[viral lotus]

Definitely couldn't hurt

[tight spade]

https://suno.com/song/310914a9-0a2d-408c-8a2e-c374464d6992 look at the promts at top then ill show you same song diffrent prompts and how diffrent they can sound

[viral lotus]

Just a thought, have you tried the prompt without reference to reggae or ska or jamaican? In other words, if you include bob marley without john ray tags, will it recognize bob marley?

[tight spade]

no i havnt but i will try right now , im certain it will just give me a generic voice

you like country / reggea

lol

it kinda works

1st gen https://suno.com/song/b45fa780-0982-4c46-8ece-b5d8e4503ed2

2nd gen https://suno.com/song/21ebdadc-a0f2-43ed-af50-34c35380c52f

i think i selected you by mistake

[ivory pawn]

Ooooo that’s sick, I gotta try getting more technical with promoting rather than just blasting the create button 😅

[tight spade]

@ivory pawn check out sunoaiwiki.com, suno.wiki. its alot more fun and alot more consistency

[viral lotus]

Not convinced that the artist names steer the generation at all

[ivory pawn]

Legend I’ll check it out

[tight spade]

im going to try removing all reggae words from song i think more so genera steers the vocals

[viral lotus]

Yes that so i was thinking, also the instruments

Have you tried experimenting with artist names outside of the genres that you specify?

Not criticizing your approach, things like this are very valuable to experiment with

And it takes more than 10 or 20 generations to figure it out of course

[teal raft]

Same one you shared with me for SDXL?

[tight spade]

defiantly genera directs the tone, here are two contrasting [GENRES: ]
[SOUNDS LIKE: reggae ]
[STYLE: melody, rhythmic, Clean]
[MOOD: happy, energetic, romantic, fun]
[VOCALS: deep male]
[ARRANGEMENT: groove, dancing, textures, Clean instrumental melodies]
[INSTRUMENTATION: ]
[TEMPO: slow, 70-90 bpm]
[PRODUCTION: Warm tones, warmth, Masterpiece, Perfectly Recorded]
[STRUCTURE: Verse, instrumental, Chorus, Verse, instrumental, Chorus, Bridge, Outro]
[DYNAMICS: energetic, builds and releases, Smooth transitions]
[EMOTIONS: Peacefulness, lively, energetic, Nostalgia]

[GENRES: ]
[SOUNDS LIKE: Bob Marley and the whalers]
[STYLE: melody, rhythmic, Clean]
[MOOD: happy, energetic, romantic, fun]
[VOCALS: deep male]
[ARRANGEMENT: groove, dancing, textures, Clean instrumental melodies]
[INSTRUMENTATION: ]
[TEMPO: slow, 70-90 bpm]
[PRODUCTION: Warm tones, warmth, Masterpiece, Perfectly Recorded]
[STRUCTURE: Verse, instrumental, Chorus, Verse, instrumental, Chorus, Bridge, Outro]
[DYNAMICS: energetic, builds and releases, Smooth transitions]
[EMOTIONS: Peacefulness, lively, energetic, Nostalgia]

https://suno.com/song/78ee6ac4-a73b-4e88-86d4-6c132fb6485e

https://suno.com/song/70a2c8ac-626c-4805-b9e4-0946d88d66fc

[tight spade]

@viral lotus well , im thinking it kinda matters, more the information about the sound your trying to go for the more consistent generation. after replacing the name i got two gens that were more reggea then nothing in "sounds like" then i tried just a name like :bob" i 1 gen was reggea second gen was country. ill keep playing around with it but im in over my credit this by 20000, from playing around

also i removed all text from "style of music box " and was surprised that in my prompts it picks it up better,

[viral moss]

Can someone write a script that removes the custom cover art people upload, and replaces it with the originally AI generated art?
I'm tired of seeing the same half naked anime women on the front page

And while I'm being demanding, for Firefox plz 🙏 or greasemonkey

[teal raft]

the nature of generative AI, unfortunately

[lofty merlin]

The don't, at least not in the direction you're expecting. Unless Suno lied to everyone, they explicitly removed all artist names from the training data, so the model has no clue what any names you enter mean. The base model may be able to try to infer, for instance, that Bob is most likely a male voice and Whitney is most likely a female voice, but it won't know what Bob Marley or Whitney Houston sound like, nor what genres they are most associated with.

[slow axle]

Anyone know why my songs are repeating the first verse over and over again all of a sudden? Used the same formatt for ages and haven't had this issue

[thorn lava]

Anyone know if there’s like a list of prompts that can me used. Not like words but like when you put in [instrumental intro] it’ll give you that. I wanna know what else I can use

[charred merlin]

Hi @leaden lark,

Finally getting a chance to catch up after a couple days of being a tourist in my home city -- always nice when you get to enjoy your home through a set of fresh eyes when family visits ❤️

To further clarify my earlier mix-up, I'm not certain that [e-mail] would be able to assist with the issue since it's largely handled by an automatic detection system.

It sounds like the issue you're running into is a 'false-positive vocal detection', where instead of being able to immediately know whether a song has vocals or not, it gives it a quick scan for anything that sounds like vocals. For instance, https://youtu.be/TAcH8TNkWOQ?t=24 it's not hard to imagine how this digeridoo could be mistaken for vocals if it wasn't reviewed by a human?

The strict voice-detection is implemented to prevent people from uploading a clip of someone else's voice and having them sing a song -- while <@&1108435550475653161> technically can't directly intervene in every instance, this is a simple manner of preventing bad-faith use of the tool, while still allowing good-faith users to use it.

That being said, I know it's not a perfect implementation, and while I'm happy to have the feature released, albeit with "safety guiderails", there definitely are some aspects of it that could use some looking into. Maybe for instances where it's legitimately a false-positive, and Suno truly mistook instruments for vocals, maybe there could be a "request review" option to have someone approve it? Even if it takes a week or more, if it's preventing people from interacting with the social media side of Suno.com, it's worth exploring the idea further in #💡┃suggestions-web ❤️

(pssssst.... In the meantime, you can try making minor adjustments to the audio upload and try to submit it again, because the system might not flag it as a false-positive each time, I've seen a lot of weird quirks.)

[charred merlin]

v3 was a notable improvement over v2, and v3.5 is just what sort of optimizations they're able to squeeze out of the v3x series while they continue to finetune further models or updates. (your feedback from ... > Report > Bugs > {whatever issue} helps make each model better, including audio quality when poor generations are flagged as a "Poor Audio" bug)

In case you haven't seen it #📢┃announcements message, tony casually namedrops "v4" like he's just popping out to the corner store for a bag of chips.

[charred merlin]

No API at the moment -- there's a lot of shady companies popping up CLAIMING to be the Suno API, and offering to provide access for a fee.

• You will most certainly get scammed by them.
• You will most certainly be breaking Terms of Service for accessing Suno, and they'd be within all rights to discontinue your access to the service.
• These companies will take your money, and scam Suno out of free accounts (likely meaning that your songs will be flagged and taken down as stolen if they're used commercially, despite whatever the scam-company tells you -- these types of behaviours also strongly come into play when you start to ask why fewer and fewer companies offer comprehensive free services -- because if you leave out a complementary candy dish for long enough, people are gonna people)

[charred merlin]

Yo @tropic fox!
In @naive quail's example, he was primarily focusing on the weather in isolation at the beginning of the song, with instrumentals being added afterward. Can you share a little bit more about what you were trying to accomplish? We've already found a Sunovator who managed to get weather noises -- I'm sure it's only a matter of time until we figure out how to nail the exact sound you're looking for 💪 🙂

[vocal meteor]

I figured as much, yeah. I'm just a developer interested in using Suno for on-the-fly music and BGM / radio. It would add something unique to a project I'm working on. :)

Thank you for the reply though!

[charred merlin]

Ooooh yeah, there's the gut-punch feeling -- my dad taught me how to play Atari 2600 and Wolfenstein 3D on his work computer (you know, typical kid stuff 😅)

I've noticed that Suno has an easier time "recalling" sound effects that are commonly found in audio recordings (sirens, bleeping, the occasional gunshot, atmospheric weather noises as we've just seen), but I'm not sure the dialup noise would be easy to prompt -- but Suno manages to surprise me literally every day.

Perhaps you'd have an easier time uploading it as an audio upload, then extending from it to create some kind of beat? As someone who relied on dial-up until WAY too late in life, I would be really interested in hearing what you come up with (admittedly, I might have to try my own, because I know dialup noises sounded different depending on the company you were dialing, so there's a very specific pattern of demon-screeches and electronic boings that "sound like home" to me 🤣)

[sinful merlin]

I made a demo yesterday without the fx and without the story laid out completely, the idea is to have the kid and the father growing together the life happens, family drama and the wholesome part would be when they call back to each other or something like that

So the idea is to have let's say three parts of the internet, said internet being involved in the father/kid split and a twist at the end

something like that

if you have ever seen the anime lain, that could be a good reference

[charred merlin]

https://discord.com/channels/1069381916492562582/1253766525458841610 There IS a known-suggestion for Suno to make their API available, but it's currently locked due to inactivity 😅

If I could bother my dear friend @unique stone to set it as active again, you could upvote the post to let the devs know that it's still something that people actively, desperately want. ❤️

I completely get that it's not a matter of not a matter of "not wanting to abide by the ToS" -- it's literally a whole new world for musicians, and there's a whole sea of brainiac developers who want to get their hands on a potentially life-changing API. It's still super early doors, so I'm sure something will be announced when they're able to handle it in a way they're satisfied with, but this kind of educated and polite input from the community can help let them know what is important to include in an API -- on a similar note, https://www.landr.com/fairai/ I was reading that LANDR is trying to add a kind of song-side API for music on their platform used for training data, so you could control song generation with increased granularity.

I only dabble with software development, and Suno is mind-blowing on its own, but I'm really excited to see what kind of "Hail Mary Touchdowns" that people like yourself will make once Suno "goes long" and passes the ball ❤️ 💪

[sinful merlin]

we should have a channel for cooking songs like a laboratoy for making songs

[unique stone]

Not sure how to determine if it's active/inactive and how to activate it again. I don't think the posts expire? They just disappear from visibility unless you search for them afaik.

[charred merlin]

Are you able to "open" it if you right click?

[charred merlin]

If you try to interact with the reactions it gives a 🚫 symbol, and you can't comment

[leaden lark]

I’ll try my best! Since the version I want to make public is my favorite version of it. I understand the need for the system, and I will try to see what I can do

[charred merlin]

It certainly "can be weird" -- I have heard of someone uploading an instrumental recording and it said "vocals", so they removed a few seconds from the recording, reuploaded it, and all was well ❤️

[charred merlin]

Looks like you got it ❤️

[unique stone]

It wasn't locked. I had to lock and unlock it again.

[leaden lark]

Thanks, I hope I can get the version after I do some tweaking on it since the way the vocals and instrumental went was my favorite gen

[charred merlin]

@unique stone It wasn't "locked", but it had expired and the post was "closed", as I understand it? Either way, I was able to leave a reaction, so I imagine others can too ❤️

[leaden lark]

Actually, if I download the song and tweak it in FL, would that work?

[charred merlin]

Yeah! I use https://www.bandlab.com/studio and https://www.bandlab.com/splitter when I just need to make a quick edit without needing to open up the whole DAW, but you're 100% on the right track 😄

If it's essentially "an honest mistake" where the AI is accidentally flagging an instrument as vocals, it might need a bit of fiddling, try shaving off a couple seconds from the beginning or end -- I'm sure you'll make the weirdest change and THEN it will work, and you'll just be left happy, but scratching your head 😛

[leaden lark]

Sweet, here’s hoping once I chop up certain bits and put them together Suno wouldn’t have the false flag again when I try to get the song

At work atm so only doing Suno in downtime/breaks, so I’ll do the fix when I get home

can I link the song? I can get a link but can’t make it public

[charred merlin]

Absolutely! You're still welcome to share those songs the same that you'd share any other, with the exception that it won't show up in searches on the social-media side of Suno.com 🥸

[leaden lark]

It’s a mix of JP and EN, that I love https://suno.com/song/d62d1d32-c482-4c7b-b6e3-ceaf28737c1f

[charred merlin]

...what a "2024 moment".

"Hey Sara, whatcha up to? You've been staring at your phone for the last 3 minutes."

"Oh yeah, sorry. I just accidentally made the coolest techno song I've ever heard because Suno didn't know what I meant by "hillbilly handpan jazz" -- I'm just putting it on Spotify and then I'll finish up."

[leaden lark]

Lmao, tbh, Suno has been cooking chill Pop songs for me lately and I love it

[charred merlin]

getting past the mind-absolutely-blown phase from ChatGPT, getting back to a productive workflow, and Suno comes along like

https://tenor.com/view/so-pissed-im-work-mad-gif-23416426

[leaden lark]

https://suno.com/song/0e516731-de73-451d-9bc7-a9a27b3440a3 my favorite song so far and it’s my most popular

I love it

[charred merlin]

😄 I'm watching Monty Don's Spanish Garden with the ever-radiant @delicate idol 💐 , probably heading to bed before TOO long, but I'll give this a listen in the morning

Thank you for sharing -- I know it's one of your more popular songs, but is there a reason that the song really resonates with you, or a certain part that you think sounds really cool? ❤️

[charred merlin]

@sweet wigeon -- I feel like you are somehow able to see inside my brain, because I have these exact same thoughts! 🤣

As Suno.com exists right now, I believe all users have access to the same tools, but not everyone knows how to use them very effectively. It would be very interesting to see something like "Advanced Mode" -- but I wonder if that would confuse some people? My mother has "Pro Mode" on her phone, which gives her MANY options that you would normally find on expensive cameras (iso, shutter speed, etc) -- she doesn't understand any of these features, but still uses "Pro Mode" because she "wants her photos to look Professional" (even though the pictures would look better on "normal" mode 😛 ❤️ )

The #📢┃announcements message Summer of Suno Volume 2 contest already has a lot of examples of people using Suno to help them learn to sing or play instruments -- I agree that it will be very exciting to see how Suno helps people grow their musical skills (whether it's by simply inspiring them to learn an instrument, or in my instance, to provide accompaniment and backing instrumentals when I'd like to "play in a band" again -- but stay in my pajamas 🤣 )

[charred merlin]

!!! Congrats!

I completely understood the sheer panic, especially considering some of the horror stories I've heard about misunderstandings throwing monkey wrenches into things, but you handled it with a proactive and collaborative mindset, and I'm sure that went a long way for Casper -- even if "it's their job", I wouldn't be surprised if the amount of less-friendly interactions they've received has increased simply as a symptom of such a monumental increase in people uploading new music.

That's actually really good customer support; I haven't heard too much about RouteNode one way or the other, but I love how detailed their response is -- "Here's what I did, Here's exactly what you need to do, Here's helpful information that might apply." 👀 I'm sure that's won you over, but do you have any other feedback, one way or the other?

[charred merlin]

Hey @mint flare, Sorry for the delay -- are you still experiencing this issue?

[lucid valve]

Hey can I ask something that isn’t related to Suno itself here?

[lucid valve]

Also, yes! I am still surprised about their custom service, I wasn’t expecting much for a free service.

[charred merlin]

If it's not tech/suno related, #🗣️┃chill-chat or #💬┃general-chat might be better suited just to try to keep conversations more tidy (but heck, I'll chat with you anywhere 🙂 )

[lucid valve]

Okay I’ll message there

[tight spade]

here is the most prompts i have put into a song , and my head hurts. but it sounds good i hope. thoughts. yes i fixed prompts and they do mostly work. https://suno.com/song/b7704fed-9e55-4c67-9048-21763a139ff4

the video show numbers that arnt there

[charred merlin]

You're right, 11pm EST -- /refresh makes Woodhouse scream it at you 😅

I'm sure you mean that you only have a single free account that you use, as not to take advantage of the free credits, but if you notice that credits haven't reset after a 24 hour wait, it's worth reaching out to [e-mail] to fill them into what's going on so they can look into the matter -- another reason why I'd love to see a "credits history" popup on https://www.suno.com/account

[mint coral]

Hi everyone! I've just start using Suno, so I'm descovering its capabilities (I'm thinking to becoming Pro next month).

[charred merlin]

...yeah, it's about time I look into making a "Sunovator" sticker for my server 💙

🥚🤕

[charred merlin]

Dear lord -- I've never needed to listen to Amyl and the Sniffers before, this is AMAZING @mystic flume yolkhead needs to open for Celtic Moonshine 🤘

[charred merlin]

ftr, I'm not an employee, but they probably don't like people admitting that

[charred merlin]

If it's still in your friend's account, can you send the link in a DM? 🙏

I didn't mention at first, but "ghost in the Suno" made me chuckle, I enjoyed it 😛

[acoustic hornet]

Makes sense. Wondering if that could be why they haven't reset. Though I've gotten no correspondence to slap my hand and call me a "bad boy."

[charred merlin]

Hrm, I've never kept an accurate watch on how many times I've played something vs the play counter, I'd be curious to know how long you need to listen to a track for it to count, if the playcount isn't aggregated until a certain point, or if there's any sort of cooldown -- I imagine they're keeping the exact "formula" intentionally vague to a degree to prevent people from trying to take advantage of it, but I've noticed that when I'm working on a song and listening to a section over and over, I'll look at the playcount later and it will say 7 or something that catches me offguard. 🤔

I wish I had a more immediate answer for you, but you've definitely got me curious as well 😛

[flint haven]

@acoustic hornet Do we have your attention now?

Jul 28, 2024, 3:12 AM

[charred merlin]

Yeah, there's some stuff that companies are legally required to tell you, and some stuff they aren't.

There's some things they're not legally required to, but still choose to in service of their users, or some things they exercise their legal right to not inform you of.

A great example is Apple not being required to let you know that your AppleCare subscription is about to expire 👀 😱 -- people get really angry when their phone breaks and they find out that their coverage ended -- technically the information was made easily available to them, it's their responsibility to make educated decisions about how they conduct themselves.

Similarly, if you think that any of your accounts have been frozen, the most direct route would be to email [e-mail] to see if they can provide any clarity. 🤞

[charred merlin]

Oh my god I missed you, you adorable geriatric. Get over here

[flint haven]

Ah it does this old man's heart good to see that someone thinks kindly of him.

Jul 28, 2024, 3:20 AM

[leaden lark]

https://suno.com/song/bb2b7c8f-700a-4ae5-8bc3-c512252ac832 so I tried to recreate the dragonforce style, opinions?

[acoustic hornet]

Hmm... I was not attacking anyone, so why the timeout? Simply pointing out that I hadn't gotten communication to note the possible infringement with a jokey way of putting it. If that kinda joke was too much, then I apologize.

[flint haven]

@acoustic hornet Perhaps sir could reread the terms of service that he agreed to when joining?

Jul 28, 2024, 3:25 AM

[acoustic hornet]

A good reply came from @charred merlin and it made perfect sense. Fair enough.

[flint haven]

I still know what you wrote....and what you did last summer.

Jul 28, 2024, 3:33 AM

[acoustic hornet]

I am aware

[charred merlin]

As far as I know, they're 100% cool with you using a single free account with a paid account (*or multiple paid accounts if you want to find a workaround to buying extra credits, but you didn't get that suggestion here and this conversation never happened 🕵️ *)

Might not be a bad idea to that the proactive approach and try to clarify the matter? I know this is a different service, different scenario, etc etc, but #📚┃suno-school message shows another user, luiz, who accidentally uploaded a free-account-song to their distributor, and didn't want it to get their other songs taken down. They took initiative to handle the matter, and all seems cool 🙂

[charred merlin]

Woodhouse is drunk again -- if you follow his rules and let him shout for a bit, eventually he'll transition back into singing showtunes.

[acoustic hornet]

Distributor? Like creating the AI tunes to make a little scratch? If so, that's not my goal with all this. I just like creating - just get a little nitpicky and frugal when things flub.

[neat nexus]

Yeah, I totally get it. When I don't like the food I'm served at a restaurant I totally dine and dash too.

Note: This is sarcasm.

[acoustic hornet]

I get the joke. Slightly similar, but if you buy a TV, Computer, etc, and it doesn't work right, do ya keep it?

[neat nexus]

Those are not consumables.

It costs every time the AI runs.

Running the AI is not free. It must be paid for.

If you use the AI to generate something, it was created and energy was expended to make that thing. It was consumed.

Or are you arguing that unpleasant food at a restaurant can be uneaten?

You do know that if you eat a certain amount that most restaurants will charge you for the entire meal. It doesn't work to eat all but a quarter of the meal and say 'Ugh, this wasn't good.'

Trying to walk out or demading not to pay at that point is theft.

Theft of services as I understand it.

[acoustic hornet]

You pose an excellent point, and it is valid to the degree of consumables. I would 100% agree if that consumable was, in some part, not made by you. Though in this context, you spent time coming up with your own lyrics, trying to perfect your structure, and the like, PART of that consumable was made by you, unlike restaurant food. Could nitpick back and forth.

[leaden lark]

Thoughts on a remake of one of my first songs? https://suno.com/song/f489a23a-92f1-4266-9bee-197b2c3d3d0b

[charred merlin]

I know there's a LOT of ambiguity regarding "ownership" when it comes to music that was made collaboratively through human-driven use of generative AI -- there's people way smarter than me trying to figure this out in lecture halls and think tanks and courtrooms and chatrooms, but at the end of the day -- it's the product of a free service, the access to which was provided with certain stipulations or caveats.

Regardless of whether it's fair or not is a whole other can of worms -- not to say it's not worth discussing, I'm just not sure this is the most appropriate time or place to delve into it.

ToS aside, I really appreciate the time and effort you put into your songs -- I want tech-chat to be a place where we can try to figure out how to work out the kinks that you feel hold things back.

That being said, even though I'm an indie-rock kinda dude, "Don't Forget" is a really awesome country tune ❤️ Violin & guitar is such a nice combination -- and I can just imagine the "crowd noise" coming through at ~2:33 when the drums stop for a second 🤘

[acoustic hornet]

Are you talking about MY "Don't Forget" ?

[charred merlin]

https://tenor.com/view/ryan-reynolds-smirk-stare-smile-gif-4318212

[acoustic hornet]

Thanks Van Milder. LoL

[leaden lark]

Rate my meme accordion yodel rap song https://suno.com/song/47990dd4-f95a-4662-9dbe-80dbf59ffaa1 thanks to someone for the cursed idea

[viral moss]

Not quite the right channel

[leaden lark]

Well I should have asked what you think of the genres in the song

[acoustic hornet]

Though it may not be the right channel, I appreciate the variance of something that is borderline polka with the accordian and rhythm next to hip-hop.

I may date myself by my terminology, but the way the chorus flowed was "tight."

[leaden lark]

I asked people what cursed thing I should make and someone said accordion yodel rap, so I thought I’d try it

[charred merlin]

(parentheses) are for backing vocals
[brackets are for metatags]

Using more than 2~3 words in metatags can "break" its recognition as a [metatag] and increase the chances of lyrical bleedthrough

[charred merlin]

lmao, this is the kind of Sunovating that I wanted to come across at 1 in the morning, not only is it a valuable datapoint, it's friggen hilarious

[charred merlin]

Feel better soon, yolkhead ❤️

[viral moss]

Haha glad I could help(?)

[charred merlin]

https://tenor.com/view/rick-moranis-are-you-the-gatekeeper-ghostbusters-horse-gif-9322927

[viral moss]

Can you stack metatags next to each other so they're executed together? Like [tag1][tag2][tag3] all in one line?

[leaden lark]

I’m debating if I should try to make meme songs

If I do something like [Build up] [tension building] that would work?

[charred merlin]

Well, looks like you helped find ANOTHER bug. This chameleon is HUNGRY for bugs.

[charred merlin]

The prompt on the community wiki are great places to start, but they are only the starting point! You can take to Suno the same way you talk to ChatGPT -- even if it doesn't always 100% understand what you mean, you can create some very interesting results simply by trying to explain your idea in different ways, because there are many different ways to explain the same idea, and Suno might imagine each version differently -- find your favourite! ❤️ 💪

[charred merlin]

There's a playlist in #1256301881966198805 that proves it doesn't work, (un)fortuntately or otherwise 🙏

[charred merlin]

There isn't really a definitive list, and a lot of it is just "suggestions" to Suno, where it doesn't necessarily parse them as "commands" or "code" like you're scripting a computer, but saying "instrumental intro" makes it think about songs where that might have been included in the descriptions.

If you have a particular effect or style you're trying to affect in a song, let us know and usually someone in this little tech-y corner can help figure it out ❤️

[charred merlin]

Juggling still applies, the more you combine all at once, the more difficult it is to ensure everything is received and delivered exactly as expected -- ymmv

[charred merlin]

Always yes 😛 Especially near the end of your credit cycle, if you only have a couple days before your credits expire, go wild and make some crazy stuff! Check out some of the suggestions on the #1252267444773326848 page, make silly songs, make intros for cool songs that you can finish once your credits refresh -- that's what I do, and where most of my experiments came from 🙂

[leaden lark]

I have 3,005 credits so I’m good for a while

[tight spade]

can you order custom credits maybe get more bang for you buck, iv spent 100$ canadian and my credits dont refresh till august thirteen

[leaden lark]

Are you a free user?

[charred merlin]

Hrm, I'm not super familiar with making club-style music, but https://discord.com/channels/1069381916492562582/1154862381407666276 is one of the most popular genre channels, maybe someone there would have a "go-to" prompt that they've tried and tested? I'm sure someone here will be able to help, but that's a good place to double your chances with some niche likeminded users 🙂 ❤️

[tight spade]

@leaden lark no im pro

[alpine venture]

https://platform.openai.com/tokenizer

[tight spade]

sry premire

[alpine venture]

this one

[leaden lark]

I only do edm every so often

[alpine venture]

passed put haha

[charred merlin]

You can purchase extra credits that don't expire between months, or you can sign up for an additional membership for whatever months you'd like to use extra credits?

[tight spade]

@charred merlin yuh thats what i did

!00$

100$

i love tinkering

[teal raft]

Thanks bud. Feeling any better?

[charred merlin]

You're talking about the build up for "the drop" in a song, right? If so, that's a common enough thing that I think people would want to get in their songs, so I can add it to my prompt experiment list whenever I get some free time to play around with those again ❤️

[leaden lark]

Yeah, for a drop

Take your time, I’m in no rush, I have a bunch of stuff to work on

[charred merlin]

I'm trying to throw all of my suggestions into #1256301881966198805 if you wanted to "follow" the post to get notified when I update it 🍻

[leaden lark]

Just followed it

[tight spade]

finally have the intensity in this song i was going for, now jut need to work on structure.... have a listen , if want . https://suno.com/song/1469cd9c-8aa5-44aa-a1a3-0417bb333927

[stiff sable]

🙋 i would also be super interested in seeing the results of this playlist experiment as its a style i'm interested in trying to make more of too soon 👀 so far my (very quick) attempts haven't worked out too well!

[viral moss]

@charred merlin I'm going through your workflow guide, found this banger 🤘
https://suno.com/song/ad23b5b8-fa3b-4f2a-86ad-89ff031eff6a

Also these experiment playlists are a godsend, thank you so much 🙏

[charred merlin]

@naive quail is to blame for these -- he got me hooked on the idea of hillbilly metal 😛

[charred merlin]

well heck, now it's 2-to-1 -- okay, you win. No one else has specifically requested WHICH playlist I tackle first whenever I get around to them, so that's as good a place to start as any 😛

[alpine venture]

lil bit, it's basicslly just a bad cold now; appreciated 👍

[teal raft]

check ur DMs BUD

Nothing super important, but neat

😏

That's two buds

[charred merlin]

This sounds mildly threatening, like there's going to be a horse head or something

🤌

[stiff sable]

something i have really been trying to do lately is do that kind of 'club' or 'mainstream' summery house music (idek what to call it so poor suno was trying to understand me) but my results weren't very fruitful. the type where you have a house beat but rather than synths or deep bass: really powerful piano chords played throughout the drop. tried combinations of piano house (which usually just makes a 120bpm song with a very luscious piano score on top), deep house etc. and trying to prompt for drops and styles in the lyrics box. so i guess its back to the laboratory for me to try and figure out what im looking for! 🧐

[opaque orbit]

Acid Trip test no. 29: https://suno.com/song/32099f2e-b3c9-4e68-b648-ccf49da82ab2

[opaque orbit]

Technically, I found this particular sound to be difficult to make. I kept getting country singers. I honestly hope they broaden the voices or I’ll just use my own

[stiff sable]

Defo hoping for selectable/save-able singers in the future! i think it's one of the most upvoted suggestions currently too so cross all your fingers!

[charred merlin]

https://discord.com/channels/1069381916492562582/1252789683176345641 harri stole the words right out of my mouth 🙂 If you follow the link and chuck a reaction onto the post, it'll let the devs know that people still are waiting for this 😛

Also one of my most highly desired features!

[opaque orbit]

Thanks!

[charred merlin]

random thought before I head to bed for the night -- as soon as this feature is added, I'm going to beg yolkhead to share these female punk vocals 😍 https://suno.com/song/9e92a99b-0cca-4b36-b38c-36dc002327eb

[harsh river]

Strange question - I made a 2-part v3 song a while back, and I recently decided to remaster it, so I extended the first part again now with a v3.5 model with the same prompt, I then put an interlude right before the bridge and I've used 90 credits at this point trying to get an interlude without an electric guitar. Note: nowhere does it say [electric guitar solo], heck, nowhere does it even describe the song as rock in the original prompt, yet I seem to only get interludes with electric guitar transforming the song into a full rock song. Is there a way to fix the problem?

[alpine venture]

oh you'll like the song um trying g to finish up

I'm convinced after these more than ever that ai music is better suited to NOT imitate existing artists as the vocal power is going to supercede them easily in terms of raw expressiveness and accuracy

[sweet wigeon]

The song would be great with natural voice - try to avoid moll chords and or darks scales to avoid that. You can use "happy chords" to get better quality voice.

[alpine venture]

and suno is in the right direction there

[tight spade]

@sweet wigeon , il try that.

[alpine venture]

like, I suspect once we have the "deep blue" of music, we'll see vocal qualities that are so good it will be almost addicting to hear

bc you get hints of it now

where it goes on a string of hitting these emotive expressive notes waaaay longer than one might have thought

[teal raft]

heck yeah I'm excited for this

At least for me, and I suspect a lot of people

Maybe not most, but maybe

That this is music now

[alpine venture]

I think once they hear it, they won't be able to help themselves

all we need is quality preference data

and we'll get there

[teal raft]

the analog hole also works for human enjoyment, not just piracy

[sweet wigeon]

Man, how do you think I have restored my music learning? I've bought a guitar again to start almost from begin. A friend should gimme keys for free so I will learn because of Suno.

[sweet wigeon]

If you just copy the text from LLama, suno will not follow it. You need to copy portions you need and add the () or [] where needed. I managed to do instrumental song using copy/paste from LLama but that was just an incident, good luck I guest.

[sweet wigeon]

Chord progression - do your own chord progression. Just learn the basics.

[tight spade]

@sweet wigeon Holy what a difference. in the song , thanks for the suggestion. ill post in a min or two. took a bit to get it right

[viral lotus]

Waiting... 😁

[tight spade]

i ran out of credits to finish the whole sond but here is half

https://suno.com/song/2d588413-4f67-4c30-ade4-cb2e7563e6f7

[viral lotus]

Not the dreaded out of credits

[tight spade]

or https://suno.com/song/da6fad37-6455-484b-a66e-fe4036d5a4ee

[viral lotus]

Very cool, I like the second version even better.

Do you know if the chord progression has any effect?

It's hard to evaluate whether the prompt is having much effect

[tight spade]

it seems to cause ill delete it and you'll notice a difference, just topped up

it sounds better with out those cords , im really limited with cords i have to ask my friend .

https://suno.com/song/64fa04be-acf9-4d45-992a-7a695e3fa8f2

@viral lotus take listen

[upbeat cliff]

Is there a command to lessen the chance of it repeating a verse instead of just flowing?

[tight spade]

structure

[naive quail]

wtf

lol

suck it, @flint haven

[tight spade]

[STRUCTURE: intro (dramatic build-up, layered instrumentation, intense), verse 1 (intense), pre-chorus (building), chorus (energetic, powerful vocals, louder vocals), instrumental break (soft), verse 2 (intense), pre-chorus (building), chorus (energetic, powerful vocals, louder vocals), bridge (building), chorus (energetic, powerful vocals, louder vocals), outro (intense)]

[upbeat cliff]

Ok I’ll flop it up some more and see where we’re at, thanks

[tight spade]

this is with cords

https://suno.com/song/effa90e5-0f0d-457f-9483-587a7eccfddf

[naive quail]

without the massive keyword/tag dump: https://suno.com/song/d74c2434-1dd8-40b7-b06c-7886cb0d8386

[wind panther]

by chance would anyone know how to get Mongolian throat singing as well as a didgeridoo instrument ? maybe not a technical question.. ? idk

[upbeat cliff]

Good lort, thank you both for the responses

[tight spade]

nice

@naive quail check this tho.

[naive quail]

I haven't tried throatr singing, but you should be able to prompt for Didgeridoo directly in the style field

[wind panther]

I did do the digeri in style field and generated 10 different versions but not one had that instrament lol

[alpine venture]

ooooooh

I'm right

I devised a test

since mBube has such a distinct effect

[naive quail]

Here's a quick test gen with a didge https://suno.com/song/dafe9059-9efc-421c-bf8d-83364f93d48a

[alpine venture]

hmmm

well, I can't tell you

hehehe

woodhouse doesn't like it

[naive quail]

throws fruit at @flint haven

[alpine venture]

basically you just repeat the last letter of the word you want to increase weight on

likely also using caps could help, since this increases the likelihood of higher tokenization on that word

but idk, I didn't test

[wind panther]

I'll try with aborig as well I was using Mongolian

[alpine venture]

but adding a ton of e's

to mbube changes the character of the prompt to distinctly mbube

despite the rest of the prompt remaining the same

I believe this has to do with tokenization, and how the LLM "considers" its response

effectively that word gets a massively increased consideration

[naive quail]

I don't think those are really necessary. I can get a didge with "Didgeridoo" as the only style. It could be the "Mongolian" and other styles are actually overriding the didge, tho.

[wind panther]

ok thank you .. I shall keep trying

[naive quail]

This one just has Didgeridoo in the style... the intro is whack https://suno.com/song/6599e571-e8dc-422d-8c7b-ec66f9dd8daa

[wind panther]

lol Wow... Wild .. interesting that's for sure

[tight spade]

soo you can re move the tags after you have completed the song, i was just using it to have consistency and control, alo to help people learn too

https://suno.com/song/85eff079-0778-403f-b6bc-959cf86c8308

[alpine venture]

ok, I'm less sure now

after more extensive testing

it did impact it in THAT prompt

but not all act in that way

hmmm

[tight spade]

ill change structure in the song and you will see a major diff

[alpine venture]

https://suno.com/song/22927e8b-756d-4f66-ad95-d052263f60bb

Still messing with it, but the numbers matter, likely bc of how it has generalized BPM

vs using other special characters

[tight spade]

https://suno.com/song/73fbd0d2-6943-4c98-acf9-da4b580bcba0, https://suno.com/song/b98fc34d-63be-460d-b1fe-dbcf5b78ab44, https://suno.com/song/9fbd9c40-f1d0-4db9-a8a9-235734712e81, https://suno.com/song/25544150-6fff-4b7a-8f19-fd7372fdd1f3

i get constastantancy when i have a good structure, its not perfect but it definatly helps

[untold briar]

Blast from the past: Using Lyrics as Melody Guidance for Instrumentals
I used to post a lot of these back in V1 and early V2, and it's easier to do now.
https://suno.com/song/3caef7cc-fc4e-44d5-bd1a-88c7d5a56c50

[charred merlin]

https://suno.com/song/65a2e520-e54a-4def-be67-d483a08a49eb

Another similar experiment — reminding myself to gather a bunch of these examples

[charred merlin]

I’m not even out of bed, and you’re my favourite person today.

[untold briar]

It's extra hard guess how or why tokenization does things in Suno, because Suno only take text/tokens as input, but it only outputs audio. (Not that it's easy to guess even in pure text to text AI models.)

But on the hand, with Suno you really don't have to understand. As long as you can hear something interesting change in the music, that's good enough to be super useful! (This is why I was attracted to Bark. Even 'bad' technical changes made the voices sound interesting so it was still fun.)

[radiant warren]

Hello! I've been having a go at making songs for a few weeks and have a fair understanding of how to prompt / use meta tags. However I have a few things I'm struggling with and thought I'd have a go asking around here. I usually make 70s/Rock/blues/oi punk styles if songs if that is important to answer any question.

1: What is the best way to write meta tags? Which one of these works better: ex. [Instrumental build-up with saxophone and drums], [Instrumental build-up, saxophone, drums] or [Instrumental build-up][Saxophone][Drums].

2: In what order does suno read and use meta tags? Are they weighted in any way? I find that it doesn’t really follow/use the tags I write, for instance if I want a saxophone solo and use the tag [saxophone solo] it doesn’t always put a sax-solo in the song. Are the tags genre dependent?

3: Is it better to use a short prompt with few genres or a longer prompt to ”narrow it down”? How are the prompts weighted, In order from left to right or in any other way? If I want that sax-solo, do I also have to put [saxophone] in the prompt or can I leave it to the meta tags?

4: Is there any way to state how long a song part should be for instance: [Guitar solo, 1 bar/2 seconds] or [guitar lick after this line] or anything like that?

5: Is there a way to force it to use a specified vocal melody in a certain part of the song like the chorus? I’ve been trying to add letters, periods, commas, dashes, spaces and stuff like that to get the melody I want without much luck.

6: Is there a way to make it pause for 1/8th note before it starts singing the second line for a chorus or something along those lines? To be able to control the flow.

7: I have a lot of trouble trying to get drumbeats that fit the music, I usually find it lacking high-hat and focus more on snare, kick drum and the occasional crash cymbal. Is there a way to get more ”rock drums”?

Maybe I’m asking too much of the tool or maybe somebody has some good pointers to help me out. I hope that I’m not too much of a hassle asking 7 questions at once. Have a nice day everybody and thanks in advance for any help!

[sweet wigeon]

1. I add the instruments into style tags and into Song name like... song name: Concert for piano and violin, Baroque style G Dur or something like that.

[untold briar]

Great questions. No simple answers, especially to questions like "what is best". Mostly you will find advice about alternative prompting strategies that sometimes work. Basically you will find more things to try. And I'm not sure Suno even knows the answers to all questions.

For example on #2. You can find Suno dev posts that say Suno tried to make the order of style tags not matter, as long as they are separated by commas. But ever since V2 it felt like order mattered somewhat to me.

[sweet wigeon]

2. and 3. - No order is important, but it seems that the count is important. If you mention viola in the title, in the style tag and three times in the lyrics meta tags, probably you'll get viola in the song.

[untold briar]

Repeating words is super super useful. In style "viola, viola" or even in lyrics [viola, viola] or [viola][viola] as a way to sort of increase the 'weight' indirectly. But not in the title. Well, of course there could be a bug. If you have an example where you are sure words in the title influence the music, post it in bugs. Not supposed to work that way for sure.

[charred merlin]

“#632”, you cheeky SOB ❤️ 🤣

[sweet wigeon]

3. you can have saxofone solo and it is important to narrow style and instruments and everything be wide as needed, but Sunno does not support structural descriptions like I have described recently. We would need some advance prompting like 10 or 20 things we need to describe. Your are on details, and Suno now is not concentrainting on details. So you want solo guitar, but you will get solo guitar with flute and tom-toms. Or guitar with tamburine and something else. You cannot limit the instruments count for some section of the song.

[untold briar]

It's hard to have specific control of instrument solos. But there are some indirect or outside-the-box ways of getting more precise control. Like the example I posted above about using widely known songs like nursery rhymes (old songs that have been covered a billion times because they have no copyrights) to guide instrumentals #📚┃suno-school message

[sweet wigeon]

"6: Is there a way to make it pause for 1/8th note before it starts singing the second line for a chorus or something along those lines? To be able to control the flow.", no there is no rhythm or melody tool developed by Suno to do that. And we need tools to do that.

[sweet wigeon]

This is cool, I'm sad I cannot send this to my neighbours 👹 .