📚┃suno-school

[amber vigil]

I have been critiquing some people in DMs - happy to do some with you - I'm hardly a professional but happy to lend my insights if you want

[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

don't really know where to post this so my apologies for the inconvenience

[untold briar]

It's fine here. Users may be able to suggest ways to get more of what you want out of 3.5.

[viral lotus]

I haven't spent much time at metatags within the lyrics, but maybe you can toy with that.

[fair adder]

i don't really know how to do that, i only know the basics: verse, chorus/pre-chorus/post-chorus, bridge, interlude. i'm not really sure of what could fit or not in the structure or what could be recognized

i even tried some songs with multiple vocals but tey overlap/switc inconveniently most of the time

[untold briar]

"big finish" as a style, and [big finish] in the lyrics, that can sometimes make a nice ending.

Oh actually I have a guess based on what you said about the chorus. But it is too long for me to explain on a phone, I will send you a reply later though.

[fair adder]

mmh i alreayd tried something such as [final chorus] or so to signify that the last chorus should be more explosive and stuff but it kept doig a soft thing xD (but that was for another song and i finaly managed to finish that one)

[fair adder]

ok then lol thank you

still i really try how v3 is unpredictible and can tweek the chords in interesting ways, especially for the bridge section

[sweet wigeon]

Can you please help me to identify what causes the noise on the begin - drums solo? It's like low frequency vibrations. It sounds like when you ride a bike where you have a metal fender, there are vibrations from the loose nut with the washer on the bolt. I used to know that sound when I was child. I prompt drums solo and you can here one of sounds like the cymbal or hi-hat creates noise. How to stop it? It is because the song is depressive? I don't like that Suno creates such depressive song when it should be uplifting. This is third part, the first part lyrics were words from folk Ukraininan song and they should definitely be uplifting. The purpose of the song was to train the words. Then I added text in 2nd extended song and there is lyrics to train the words for the song but there is something with child, illness and death. So Suno changed the mood to sad. But I here these terrible noises and have no idea how to prevent it to be created. It also can generate when synths are used (long, deep notes of synth keys). I just hate it. It should be uplipting Irish-Celtic style and this is the result. https://suno.com/song/cacb2b82-ae49-4c52-9ffa-07508b8cc65f Would it be possible to specify different drum instruments to prevent the noise? Or different drum technique?

[sweet wigeon]

[Epic] or [Epic bridge] and on final [Epic outro] or [Epic fadeout]

[fair adder]

ok i'll try that, thank you

[graceful condor]

Other than extending a song, is there any way to get a certain vocal that I have heard in my Suno library page?

[naive quail]

I encounter that effect a lot when prompting for darker music. Once it Creeps into your track as a constant background sound, you're probably stuck with it. You can often extend from a point in time before it starts to steer around it. Unfortunately, in the case of that track, it's pretty much from the beginning. I did a couple generations using your prompt and didn't notice that effect in any of the other generations. Are you getting it with every generation for that song?

Jump to 02:50 in this song to hear it even more pronounced: https://suno.com/song/14629718-64c3-48ef-9493-ffc78cb99b47

Unrelated to that (probably), Suno doesn't currently have any concept of negatives. Having things like "no synths" and "no hi-hats" is more likely to actually GIVE you those things. It IS a requested feature, tho... so go vote on it! #1253341197917884530

[astral flame]

did the outage change how suno is operating? the same prompts ive been using are resulting in very strange results

[astral flame]

wow. all songs being made are trash right now! nothing like what i have been making with the exact same prompts in the past!

[charred merlin]

https://discord.com/channels/1069381916492562582/1259902708765167709

[charred merlin]

No.

[sweet wigeon]

"Are you getting it with every generation for that song?" I think yes, it's from the original song where the dark words were present. The problem is that we don't know how to call this effect so if negative words would be implemented we actually don't know how to prompt it.

[naive quail]

Let's name it "Dave"

I look forward to the day that "Not Dave" actually works in my prompts. 😄

hrrm... interestingly, the track I linked actually HAS "Not Dave" in the style prompt.

[untold briar]

I don't know if it's planned but l suggest that and much more context manipulation myself, often. 🤞

[modern obsidian]

guys, how possible we have the possibility to re-use a prompt but not a seed?

so we can get a similar melody to the previous generated?

.<

that would be a useful thing instead of the "new super useless cool look"

:|

[untold briar]

Check the mini polls https://discord.com/channels/1069381916492562582/1163842729231528047 option 🇧 "remix/cover" sounds a bit like that

[forest rampart]

idk but i just found this

looks like image is missing xd

[desert wren]

why is the style prompt being manipulated after creation?

[wet minnow]

you are in "non custom" mode. this modes takes the input and transforms it into styles and lyrics. use custom mode if you want to be more specific with styles and text. ✌️

[sweet wigeon]

Oh no! OMG. I have identified the sound! ChatGPT is great. https://www.youtube.com/watch?v=awsOCjLbh_Q

[naive quail]

I'm soooooo making 5 of those!

[sweet wigeon]

They call it Remo Spring drum. It was AI who game me advice, so I have written it down on my blog https://hudebni-vychova.blogspot.com/2024/07/identifying-round-metal-vibration-noise.html a and this is the Remo Spring Drum here: https://www.youtube.com/watch?v=K91cdbmCCfs

[naive quail]

I have some mini remo bongos around here somewhere... and some floppy springs. To to break out the hole cutter 😄

[sweet wigeon]

We could ask Suno to retrain and retune the model so that is never uses this noisy effect until specifically named in style tags.

[naive quail]

I don't think it'll be that simple. Suno's probably not creating that specifically from spring drums but the effect is very definitely similar.

This little bugger is going to become a spring drum

[sweet wigeon]

It's huge

Ok I created a song with REMO spring drums. https://suno.com/song/819aceb7-3d18-47c7-b4b3-99920e9aac17

[half shard]

Here is the second song from our AI musical “The Blacklist” (from the TV series). Unfortunately, as soon as you choose the musical or Broadway genre, the voices become overwhelming, which gradually demotivates us to continue working on the project. Everything goes great with pop and rock, but as soon as it becomes classic, the AI ​​fails.

Rollen in dem Song:
Agent Donald Ressler, Agent Harold Cooper
Chorus: FBI Team
Style: Classical, Waltz

-> https://suno.com/song/e9a9fc9e-0d85-4c54-a1ed-fa2635785cb4

[sweet wigeon]

What did I wrong that the song I posted is not playable here on Discord?

[half shard]

The song starts with a 50-second instrumental, which we first created via Suno and then uploaded as audio in WAV.

[amber moss]

just a small question, is it possible to make a suno song pan the lyrics from left to right?

[sweet wigeon]

I love it.

[half shard]

Thank you very much. 🙂 What annoys me, however, is the clipping and the cracking when the voices get louder (in the chorus). This makes the song unusable. Suno just can't do Broadway and musicals.

[sweet wigeon]

I see now. It sounds like Vinyl record scratch. They did mistake when copied Vinyl record, I think.

[half shard]

Yes, that is to be feared. However, it is the same with all musicals and Broadway songs about Suno. Even when I do a modern piece, in a rock style like Dance of the Vampires. As soon as the style says Broadway or Musical, Suno just produces crap.

I also noticed when you load Suno's songs into Audacity that both channels (left and right) are always the same. Apparently Suno only produces the songs in mono. There are no differences in the channels no matter which piece I use.

[cold zenith]

what's the best text to video for making a video clip ? except noisee ?

[cold zenith]

i choose only instrumental to finish my music, but it take some previous lyrics anyways. how to avoid that ?

[half shard]

Now I'm happy. I redid the song completely and after almost 10 attempts I got the result I wanted. And Sono even took the different roles into account (Ressler, Cooper and Lizzy). I simply left out Musical, Broadway and used Chanson, Voice Ensemble as the genre.

-> https://suno.com/song/fd5da3a8-866f-47da-b0b4-6123b8efebc3

Now I just have to see how I can put an instrumental in before the vocals. I'll probably have to edit it with Audacity.

[upper echo]

does someone have a good way to have only one instrument ? Suno seem to love suddenly add percussion or random instrument

[untold briar]

I've noticed this too. I'm usually not looking for truly just one instrument and not bothered by it though. I suspect there are good prompt tweaks that work, but I haven't tested anything, I'll keep an eye out for some.

[stark vessel]

[Solo Flute] [Mono Instrumental] smthn like that

Anyone know if WAV files are still working? (Im purely mobile/ios) every time I press it for any song, it doesn’t do anything (unlike mp3, a “do you want to download (song)?” Will come up

[untold briar]

This can be difficult. The simplest thing to try is repetition "instrumental, instrumental, instrumental". Also in lyrics [instrumental]. Other words common in instrumentals like "solo" could be worth a try.

If this doesn't work, the only really reliable way is to use multiple Extensions and don't use Get Whole Song, so you remove some of the lyrics from Suno's memory. Then you wait until instrumental starts in a new clip, Then Get Whole Song on that, and finally extend again to finish the instrumental. I may not have explained that clearly, but about to step out. It's sort of like advice to do a style/genre change in a song, if you look for messages explaining that.

[untold briar]

WAVS for sure work on desktop, but there may be a mobile bug, off the top of my head not sure. What if you try 'desktop' mode in Safari browser?

[cold zenith]

I just made a song and from the beginning the ai, goes to the last verse.....

[untold briar]

This can happen even with a perfect prompt. AI just wants to do it's own thing sometimes.

However, this happens more often when there is lots of repeated phrases in a song, it can 'jump' to the wrong phrase in a new verse. You can also try numbering [verse 2] [verse 3] etc.

[stark vessel]

Same thing, nothing popped up

[stark vessel]

WAIT NO IT DID IT- Weirdly enough, the song had to be playing and you gotta press the “…” there 💀

[hushed nymph]

is there a way to make Suno follow the actual lyrics layout and prompts IN ORDER and not repeat random parts? had several generations in a row repeating chorus and first verse over and over and not going to the rest, and another just used second first twice, and skipped to the outro and then back to second verse, never making the 3rd verse or bridge... its extremely frustrating to format as much as possible and then suno just ignores it and you get half (or less) the song you want, x3. Another song the instrumental went until 3:30 BEFORE the first verse out of 3 (and chorus) started. I feel like its trolling for money/extra regens and it's failing and wasting my time most tries. can't there be a setting to keep it closer to formatted layout?

[untold briar]

There is no simple answer. But as general advice, when you Extend a song and there is a problem like Suno repeating the wrong verse, the most important thing to change is the Extend From Time. Keep backing up 5 or 10 seconds. This usually means you have to redo some lyrics, but it can also avoid problems like that.

[hushed nymph]

i didn't extend at all, its just wasting original first-generations on repeating the same and wrong parts 😦

[stark vessel]

Can you copy and paste your lyrics here? (Exactly as it is in suno)

[modest coral]

newest error?

[hollow hare]

If it was reproducible than it would have been likely. However, it seems it was a coincidence (likely a hallucination from how the AI expected a certain pattern to go?)

I did a Reuse Prompt and the lyrics seemed fine at the beginning

• https://suno.com/song/e1af35d1-1d52-447b-ac95-28aabfd2893f
• https://suno.com/song/f89ffe4a-1387-467d-8d70-5388eead0ba6

[hushed nymph]

[instrumental]

[intro: emotional]
+lyrics, 6 lines

[Verse 1]
+lyrics, 4 lines

[chorus: emotional]
+lyrics, 6 lines

[Verse 2]
+lyrics, 4 lines

[instrumental]
[chorus]

[refrain: softer]
+lyrics 6 lines

[instrumental]

[Verse 3]
+lyrics, 4 lines

[instrumental]
[chorus]

[Bridge: slow rap]
+7 lines lyrics

[instrumental]
[chorus]

[outro: wistful]
+4 lines

[refrain: slow]
+6 lines

+2 random lines

[chorus]
[instrumental]

+4 lines

[instrumental]
[coda]
[end]

now this new generation is skipping chorus

[stark vessel]

Try making sure everything is capitalized, and [Slow Intro] [Slow Chorus] [Soft Refrain]

Other than that, maybe just generate only HALF of that

[hushed nymph]

and put the other half in an extend?

[stark vessel]

It looks like a lot, so in my experience, suno will try and squeeze everything tight to fit into the song

[stark vessel]

Yes! Try that

[hushed nymph]

ok i didn't know capitalizing was important

[stark vessel]

Not sure if it is or isnt, but it seems to work for me, that and the cutting things in half- especially with large amounts of stuff to put in the song

Lmk of it works ehhehhe

[hushed nymph]

alright, i'll try to tighten it up too

[untold briar]

It's normal for that to happen once in awhile, but if it seems to happen almost every time, there might actually be something in the prompt that is confusing Suno. I see others are helping you already, feel free to message me later when I'm online again as well. I will just say that prompt is quite long, maybe try cutting it in half?

[hushed nymph]

I will be trying that soon, thanks

[untold briar]

In particular, cut the prompt before it repeats a chorus, or any verse 100% repeated. Repeated blocks of text, in general, Suno sometimes jumps to the wrong one.

Remember if you clip is short you can always Get Whole Song to 'put it back together again' before extending, to make sure Suno remembers the last chorus, when doing an extension where you want to repeat it.

[hushed nymph]

ahhh, i see the method now

[sweet wigeon]

I liked classical instruments like violin, flute, pizzicato violoncello and the waltz counting. That was great in the previous version. I have to try Waltz and classical music one day.

[hushed nymph]

now i have it skipping chorus and then a 32 second repeat of only the refrain and nothing else.... I think i need a break before I cry

[jovial hearth]

quick question, can suno read International Phonetic alphabet?

[daring lava]

Hey @jovial hearth 👋 thanks for reaching out! Currently Suno does not support the International Phonetic Alphabet (IPA) for generating songs

[jovial hearth]

ah, shame

will be hard to try to get my conlang to work in it

[daring lava]

It’s a great suggestion. If you’d like to, maybe add it to the #💡┃suggestions-web channel

[safe hazel]

If you have a string of extensions for a song (like Part 2, and Part 3 etc) and you only want Part 2 and Part 3 (but not part 1) is there a way to do that? "Get whole song" will include all parts.

[viral lotus]

Does anyone know how to permanently delete a playlist that is in trash? My playlists in trash are showing in the "add to playlist" box, which is not helpful.

[lament swallow]

200 credits dead ?

@charred merlin

[worthy phoenix]

How i can assure that the AI read properly when I ask to create a song with woman singer?

[safe hazel]

Use "Custom mode" not the simplified mode

[worthy phoenix]

Yeah, i always use that mode

[safe hazel]

I will usually put "Female Vocals" in at least twice. Maybe "Clear Female Vocals, Moody Female Vocals" something like that. It seems to react to the number of times your include it in the style prompt

[worthy phoenix]

I put "woman voice", maybe is hardly to read that thing for the AI.

[safe hazel]

Try "Female Vocals" I rarerly get the wrong vocals when I do that

[worthy phoenix]

Well! Thanks for the answer, will try that prompt next time

[static bolt]

Hey everyone, how do I make the male lead vocals and female back vocals?

[rose jasper]

If you specify male voice, female voice in the "Style of Music" and then when you are going to be prompting it in the specific part of the song write down [Female Voice] & [Male Voice] give it a few renditions and then upvote on the ones that it works on, and Suno will eventually "get it"

[robust ermine]

Tell me I’m not going insane! One version is rendered with v3.5, another one with v3 (not telling which one is which). Tell me which one sounds better, especially vocals (vocals start around 0:35)

[wicked mortar]

I'm leaning to (1) being v3.5

[rose jasper]

Try to add more emphasize on songs, by adding the following behind the words:
~ = Whaling
! = Emphazise

• = Mention it loudly
  ? = Question based

[robust ermine]

It sounds better doesn’t it? Guess what, it’s v3! I made the entire song with 3.5 but then re-rendered it with v3, and I like the result so much more.

(To rerender, I continue the entire song with instrumental → “short silence” as the prompt, and choosing the different model, then getting the whole song, now with the new model.)

(It’s also free because the continuation is usually around 2 seconds)

[lament swallow]

Can I get a refund for my 200 credits? No one responded to my issue.

[warped furnace]

I think it would be best if you contact the billing department for that kind of complaint.
[e-mail]

[sturdy crescent]

is there a list of all the instructions you can give and a syntax of them?

has anyone had any luck giving instructions for pauses or any kind of downbeat type "land on downbeat" etc with custom lyrics?

[pliant pumice]

How sometimes song are no long 4:00 minuts but really more minus?

[hollow hare]

That is the maximum time a generation can be made in v3.5, it is not something that the AI actively aims for.

Usually the song ends earlier when it thinks it reached a natural end with the prompts it was given and fitted.

To increase the duration, you can

• slow down the tempo (e.g. use slow tempo in the style prompt)
• add more lyrics
• use instrumental tags to add small musical delays, etc.

[hollow hare]

Check out the entries at the unofficial wiki https://suno.wiki and other links listed at #1151542418991239168

There are many more unknown prompts waiting to be discovered. We do not know what exactly the AI learned but it can be quite surprisingly flexible in interpretation so do not be afraid to experiment.

[mystic flume]

Did you try your re-render technique with 3.5 as well?

[mystic flume]

You want a refund for $2.00?

I'll venmo you $2.00.

[lament swallow]

Refund credit, no money

[lament swallow]

Yes, and ?

[mystic flume]

Am I missing something? You want a refund for 200 credits and I offered to send you enough money to buy 250 credits.

[lament swallow]

I cannot receive money.

[sacred bramble]

So, how to avoid repeated verses? Is there any structure that can reduce it? Or a certain number of lines?

[hollow hare]

Someone made an advice post for that issue in #1252267444773326848 : https://discord.com/channels/1069381916492562582/1261146873331318835

[lament swallow]

I have a lot of music that speeds up during rendering; it can double or even triple the BPM in a few seconds.

[sacred bramble]

Ty

[mystic flume]

Tinda's Railroad Analogy

This is an overly simplistic way to describe the process, but it's serviceable enough and makes sense to me.

Think of Suno like a railroad.

When you generate a song, Suno builds tracks pieces from your input, especially the lyrics; straightaways, curves, tunnels, bridges, etc. It will then use these pieces to get from A (silence) to B (a banger), based on a random seed and its interpretation of your lyrics, structure tags, and style prompt as a whole.

These things together create a topographical map, so to speak, which is the terrain the railroad needs to cross. Suno will always stick to the simplest or most efficient route to get a result it thinks will sound good for each individual generation.

If the track pieces (your lyrics and structure) match up naturally with that simple and efficient route, then things will go smoothly and you'll get a good result immediately.

But if you put a straightaways (chorus) where it needs a bridge (bridge), for example, it will have to build the bridge itself using the pieces you gave it.

Likewise, when you're trying to extend a song to fix or change something, the spot (timecode) you are branching off from has a lot to do with how easy it will be. If I'm trying to add a tunnel (like, say a guitar solo) but the part of the "map" I'm attacking doesn't have a mountain to "tunnel" through, Suno will balk.

There are also hiccups and hallucinations too, like with any of these AI models.

I find some generations seem to be more forgiving than others, too... which to me feels like I got more of a flat plains topographical map so Suno is happy to take pretty much any route I ask it to, whereas other times I'm trying to tunnel through the Rocky mountains.

Anyway that's the analogy. It's simple, but I think it's good. Happy songwriting! 💜

🚂

[fair adder]

i'm actually very interested by this technic but i don't think i got it... could you please explain it again?

[foggy wing]

I'm having some issues where tags in the lyrics box alters the sound of the song completely. Say for instance I want female lyrics on the chorus and male on the verses. If I add it like this "[verse 1] (female lead) lyrics" I sometimes get the desired effect without altering the style of music. But sometimes it is reversed and sometimes I get no effect at all. I have tried "[verse 1] [female lead] lyrics" which messes stuff up badly. However "[verse 1 (female lead) lyrics]" <--- notice the changed end bracket, can do wonders, but also change the sound of drums, guitars etc to something completely out of style. I wish this was more predictable, I like the style I have found, but maintaining the style makes for very hard randomness in generating the vocals. I am sick and tired of finally finding the perfect chorus, but then the verses are all wrong.

[foggy wing]

Also... I can see clear effect on adding tempo and rythm to the tags box, like "4/4, 120bpm" versus "3/3, 140bpm", but when I add key and scale like "D-major, dorian" or "C-minor, aeolian" I experience a lot more randomness. Does anyone know the best way to instruct Suno on this? Would it be more successful entering key and scale in the lyrics field instead? I have burned over 1000 credits on just one song now trying to find this out.

[cunning rock]

If it’s possible, could we include more genres and references to accurately generate what we request?

[foggy wing]

What are you looking for? Genres are usually pretty accurate, but subgenres are usually defined by energy and intensity. "Heavy metal" will give a certain sound, but add "Heavy metal, intense energy" and you will have a completely different subgenre. "Heavy metal, intense energy, technical riffs" will be even more defined to a certain band.

[long cargo]

I have a question, I have made songs where it was written in Japanese and the pronunciation was correct, but if I do it again NOW with the same songs it just comes out in English, does anyone know how to make it Japanese again?

[cold zenith]

try again it might be a bug

[cold zenith]

are you saying that there's major difference between : " [verse 1] (female lead) , [verse 1] [female lead] lyrics and [verse 1 (female lead) lyrics] ".

[charred merlin]

(parentheses) are predominantly backing vocals

[metatags] are “localized” style prompt injections that work in tandem with, and influence the “style” section.

[sudden zealot]

/chirp

[flint haven]

The /chirp command is no longer working

Chirp is retired, and no longer on discord. You can make music now on our website: https://suno.com

[cunning rock]

The genres in looking for are more ethnic and ancient sounding, such as Ancient Greek, or Vietnamese bolero, traditional chinese, etc

[oak meadow]

Is it me or does V3.5's music seem uninspired and repetitive? V2 made far better beats, but lack quality audio, but when it comes to 3.5 everything sounds the same and it only seems to use the same 4 instruments. Asking for a friend 😅

[sweet wigeon]

Why do you needed? Does Suno incorrectly pronounce some language? I used more languages like ancient Hebrew, Ukraininan, Czech, Japanese and Arabic and I did not find much problem with pronunciation. Of sure I could want to make Suno to pronounce differently because languages develop in time, and modern pronunciation may not fit ancient time song.

[foggy wing]

Incorporation of IPA phonetic alphabet would be an extreme lift for Suno.

[alpine crow]

👋 Hi guys, can i check with Suno's Bark AI model, is it possible to generate a TTS voice by passing an emotion inside a bracket? ie: [happy] What a wonderful day! would sound excited, different with [angry] What a wonderful day! would sound sarcastic

or is there a special prompt to trigger generation?

[charred merlin]

@untold briar is our resident Bark expert, as this Discord is primarily focused on Suno.com (the evolved form of Chirp) -- though I'm not sure to what degree he's back up and running after a serious power outage.

I believe he's accepting to DMs on the matter if you have patience and wait for him to get back to you.

[alpine crow]

okay, thanks so much for the pointer @charred merlin I'll DM him.

[trail badger]

Currently the model is producing very good results. If any action has been taken, please do not touch it.

[charred merlin]

The model is not fiddled with in the way you repeatedly suggest, there aren't many ways to be more clear regarding this.

[trail badger]

I'm not saying it's already been done. "If it's been done," I said.

Let's get this straight. I am not explicitly saying that action has been taken on the model. The reason why I inform you is that if a change has been made regarding the model, server, etc., the "Developers" can understand through the users whether the model is working properly or not. We cannot know what actions have been taken or not regarding the model. Users only provide information.

[merry barn]

does anyone know how to change the picture generated with the song?

[wicked mortar]

the three dots on the right of the song, then edit details, but you can only upload a new one, not generate one https://i.gyazo.com/eba05394dd7c61d15d61b63f321deb42.png

[naive quail]

Should be able to generate a new one, also.

[wicked mortar]

if there is, I'm not seeing it https://i.gyazo.com/9bbcfa5bce8d5f8ee1fb87d8377147c9.png

[charred merlin]

I love you dude

[naive quail]

lols @charred merlin

[wicked mortar]

ignore me....... under add image, you add prompts face palms

[naive quail]

that was to the second 😄

[charred merlin]

gtf out of my head

[naive quail]

hahah

[charred merlin]

oh man I needed to smile like that 🙂

[naive quail]

I still kinda want to make that dead rat pic the cover 😄

[wicked mortar]

tbh I've never felt like the image generated with the song, was something I felt the need to change, so I never looked

[naive quail]

You must have missed the Bird Days

[charred merlin]

I miss mutated guitarist birds -- but "secret agent dog" or "enraged banana" are pretty hard to top

[naive quail]

every image was a weird deformed bird holding an instrument.

heh

[wicked mortar]

ok, that's sounds funny, we got any examples, cause I need to see that

[merry barn]

thanks guys

[naive quail]

someone will have to dig. I've purged my library down to 4 or 5 pages.

[charred merlin]

legit thankful that franklin added page search by number -- I'm refusing to believe I have this many pages

[naive quail]

still have some in the trash

[naive quail]

my trash:

[naive quail]

I am not missing the birds.

https://suno.com/song/aa5f4ff2-e9bd-4205-8bfc-5bbd58621989

[charred merlin]

Lindsey has called them byurds a few times because of this, btw, the adorable little beh-behs.

[naive quail]

Thank Ze Frank for that.

https://www.youtube.com/watch?v=Biagyb7AcK8

[lament swallow]

Again... 48s.... 53s.... hum...

[silk mulch]

I don't know what's going on but lately all of my sons have been going up up to 4 minutes even though that they cut off earlier and I'm ending up with a lot of empty space in my songs

[hollow hare]

v3.5 is known to have trouble ending- it probably listened to a lot of songs asking the DJ to keep the music going 😛

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)

[hollow hare]

Can you link me to the song? We can't help you without more info about what prompts and lyrics you used. You can DM me the link if you do not want to make it public.
(This is both to help avoid cluttering this channel with the lyrics you used and mainly makes it easier for Reuse Prompt to test some alternative prompts)

[lament swallow]

No lyrics, it's instrumental

[calm vector]

After creating a song I can't remix the song as it is without changing the mood, the base, the voice..? For example, if I make a mistake in writing, how can I reshape the song?

The Mood that generated me is beautiful and I would like to create more songs with the same voice and base... can't we take it from that? Or is it just not possible?

[naive quail]

Unrelated to the short tracks, which I can't reproduce (I even got one that was 03:59 with that prompt), you may want to try the english spelling for "violin" if that's the instrument that you're trying to get. I don't think Suno recognizes the french spelling.

[naive quail]

You can use the Extend feature and pick a point in time that's just a few seconds before the mistake. You'll probably need to clear out the lyrics/tags that have already been performed to give Suno the best chance of picking up where it left off.

[calm vector]

The first part of the song why can't I change it?🥲

I would like the Mood of my track to always be the same, by default, and for the artificial intelligence to remind me of the voice and the Mood I want, exactly like a song

I am very happy with the site, and I was going to pay the monthly subscription if it allows me to do so, I want my songs to have the exact same Mood, the hashtags are too generic..

[naive quail]

We don't currently have a way to pick the same voice, but you should be able to get a similar feel by using the same styles in the Style box when Custom mode is enabled. I'm not at all familiar with the basic mode (Custom toggle off); it may be different.

[hollow hare]

I found that if you are looking to make a pure instrumental song, it is better using various instrumental tags in the lyric box in Custom Mode rather than relying on the instrumental switch.

It mostly helps avoid hallucination (but this still depends significantly on the genre used) and can pad out to the max time if you have a sufficient number of it (as well as giving you control on how/what you want)

Shiroken had noted (https://discord.com/channels/1069381916492562582/1261146873331318835) that three metatags can be used consecutively without a break between them

Example:

• RayanAIR's Novo Lands

Here is a generation of your promots with my generic instrumental tags:

• https://suno.com/song/974809fa-81c2-4271-8ef4-ff9609a4433b
• https://suno.com/song/34640a52-af9b-46be-af87-0b5a58a50e27

[charred merlin]

Hey all -- sorry for dozing off for so long, the pills I usually steal from Woodhouse's medicine cabinet must have expired.

Or he started swapping them out with tic tacs.
-# They did taste like orange...

Gonna start working in reverse order through backlog, so if you're upset that new questions are getting answered first, just politely ping and I'll bump your spot ❤️

[naive quail]

hides his orange tic tacs

[hollow hare]

No problem, we all need a break now and then. And you should know better than getting into Woodhouse stash 😱

[calm vector]

Okay, thanks anyway, it's a real shame in my opinion, if a person wants to create a "singing character" how do you choose the voice or the Mood?

I'm too sorry he's not there

Maybe it can be of help or feedback

[naive quail]

I'd choose the mood by populating the Styles box in Custom mode with the styles that I want. As for the voices... it's going to be a crapshoot right now. You CAN fairly reliably get male or female voices by specifying them in the Stlye box and you might have some luck with additional descriptors like "Gruff Male Vocals" or "Soft Female Vocals".

[hollow hare]

There is a poll going on for future development at #🗳┃mini-polls message - Option C sounds like what you want 🙂

[calm vector]

Thanks

[charred merlin]

https://discord.com/channels/1069381916492562582/1252789683176345641 Hoo boy is this one of the more popular requests (upvote it to show how important it is, but I think the poll shows that it's on their radar!!!) -- while it's hard to guarantee a similar singer between generations, it can help by using a similar external vocal source (you singing, song generation websites, Suno's open-sourced text-to-speech model) and using the audio upload to try to "suggest" to Suno what type of vocal waveforms to use as context & continue from.

Beyond that, I've seen some interesting tips (on getting the same intonations and vocal timbre, not simply getting male or female), but nothing that's jumped out as more than wishful thinking or confirmation bias imo ❤️

[charred merlin]

you're a crapshoot right now -- gimme those tic tacs mans stop hogging them

[naive quail]

lol

"Chain-smoking Grandma Vocals" was ineffective, btw.

[calm vector]

If you add this option to reuse the same voice and backing track, Mood, I will purchase the annual lifetime subscription whatever the cost

Thank you with all my heart 🤍

[charred merlin]

I know -- it's... I can't even believe that this just sprang up out of nowhere, like a musical dandelion headbutting it's way through the schoolyard asphalt. NONE of this was possible just over a year ago.

[charred merlin]

Now that I think more about the better-handled implementations of consistent-leaning vox, @mystic flume might be a good person for you to attempt to summon for conversation -- she's created an AI band "Celtic Moonshine" that seems to handle this fairly well (or at least to my ears).

I had to take the opportunity to share this video again because she makes her own visualizers for her tracks, and I don't know if I've ever seen her re-use the same effect twice -- just as a subtle means of flexing her tech-knowhow, like it's no big deal to just constantly innovate
-# you're making us look bad, just take a couple years off so we can catch up for a week 😛 💜

[lament swallow]

thx. Any idea why the music speeds up almost all the time?

[naive quail]

I noticed that in a couple gens when I tried the prompt you supplied. It may be related to "Electro House" but I did a few more after changing Violon to Violin; it didn't happen with them. It seems pretty clear that Suno doesn't recognize "Violon" so it could be Suno's attempt at trying to do.. something with it. It could have also just been a coincidence, tho... that was a pretty small sampling.

[lament swallow]

Yes, Electro

[charred merlin]

None of the songs I produce through Suno seem to have this issue, leading me to believe it's a matter of the prompts you're using?

Can you share some songs where the "speed up" is particularly noticeable (as well as a timestamp of when the change in tempo occurs) for a deeper look?

[lament swallow]

The problem mainly comes from electro, house, techno, hardstyle, and similar genres (anything related to electronic music).

[charred merlin]

Again, I've seen it happen once or twice when I've generated similar styles, but not enough for me to consider it to be a "widespread problem" -- if you're able to compile a list of songs to demonstrate this, I might be able to look a bit further and try to see if I can shed some further light, similar to the advice that @naive quail had given just above, but without seeing a direct example for comparison, or more collaborative exploration of the issue, there's too large a degree of speculation to give actionable advice beyond "try to experiment, see what works for others".

[naive quail]

here's an example using their prompt https://suno.com/song/01fbec1b-2923-4010-b5e8-ba4c50b4fe30 . I think it's related to the style

[lament swallow]

Speed ahahahahha

[vital plaza]

Thia is pretty intaresting xconcept you can make a really serious concep out of this... this has some good style foundation(Comedy-House) Tex- groove breaks- sound like awesome concept to build

[naive quail]

I got that in a few with "Violon" in there, but none of the ones with "Violin", so I'm not really sure which one is pulling that in. Maybe the combo of a word Suno doesn't know with a particular style

[mystic flume]

Oh, there's no magic bullet for getting the same voice. I use the same voice prompt every time, which helps, but mostly I just keep trying until I get the right voice. I've had to trash a bunch of really good interpretations because the voice was wrong. So, it's really just a matter of tossing credits at it until I get a banger that has my imaginary lead singer's voice.

[naive quail]

Yea, I had some fun with that. 😄

[lucid valve]

Hey!
I've created my album using Suno AI Pro, am I allowed to distribute my music on all streaming plataforms such as Spotify, Amazon Music and what not?

Also if I do, do I need to contact Suno about it?

[mystic flume]

Yep. It's your music and you can do whatever you want with it. Suno relinquishes all rights to you if you are a subscriber.

[naive quail]

So, it's really just a matter of tossing credits at it - I know this song well!

[lucid valve]

Commercial Use: Subject to the Content Section below, unless otherwise expressly authorized herein or in the Service, you agree not to display, distribute, license, perform, publish, reproduce, duplicate, copy, create derivative works from, modify, sell, resell, grant access to, transfer, or otherwise use or exploit any portion of the Service, and any Output, for any commercial purposes.

I found this in the terms of service

[mystic flume]

"Subject to the Content Section below" means the Content Section governs the exceptions which are "expressly authorized herein"

[hollow hare]

So long as you generated those songs while you had a subscription, you can monetize those songs since you were granted ownership.

Harber and I did some talking about DistroKid at #📚┃suno-school message
That post also has my general blurb about distributors too

TL;DR: Read the TOS of the distributors, especially in regards to Royalty payment. Avoid TuneCore and cdbaby since they do not like accepting AI songs (and TuneCore has some very suspicious terms regarding their royalties which Top Music Attorney on YouTube covered)

[vital plaza]

no! If you on Paid subscription you own all the tracks

[lucid valve]

Also I really need to get this off my back... My Suno PRO expired and the distributor still haven't accept it, do I own the rights forever or just until my subcription end?

[mystic flume]

You own all rights to songs you made while your subscription was active, in perpetuity.

[lucid valve]

So I don't need to wory about losing the rights?

By the way, is RouteNote a reliable distributor?

[mystic flume]

Correct. Suno relinquishes all claims they might make. That doesn't stop somebody else from challenging it, though. Like if your song sounds too much like "Beat It" or you used plagiarized lyrics, etc. But it won't be Suno that sues you.

[vital plaza]

Use distrokid. They will distribute it on Apple, and spotify/ it takes up to 24 hours. singles takes a few hours. I higly recommend distrokid.com they accept AI music

[mystic flume]

Suno = No Sue? Heh... hehehe.... hooooo

[charred merlin]

If you're questioned about it, the only current way to "prove ownership" would be to show screenshots that show your account was active at the time the song was generated -- couldn't hurt to include the parts that created the "Full Song" if you need to prove that (GOOD REASON NOT TO TRASH THEM)

[charred merlin]

I'm not familiar with them, and would need to research the matter further to give an honest unbiased opinion.

My wholly biased opinion is that DistroKid is the best way for newcomers to dip their toes into the music industry for the lowest cost, while retaining confidence that your AI generated music will remain hosted.

[lucid valve]

I was looking for a free one

Okay guys, thank you all so much for the help!

[mystic flume]

I use LANDR because I really like their AI Mastering. They've been really good about distributing my music as well. It's not super expensive either.

[charred merlin]

I was just recommended one, gimme a sec

[mystic flume]

You want a free distributor? That seems unlikely to exist. But if you find one, let us know.

[lucid valve]

Okay, I'll be waiting

[lucid valve]

RouteNoute is free

But they take a LONGGGG while to accept your realease

[charred merlin]

https://www.unchainedmusic.io/

[charred merlin]

That would be the comeback to the whole thing. DistroKid is around $30CAD PER YEAR, plus they handle so much stuff that you need if you want to get started easily (youtube music account, spotify, itunes, blah blah blah)

[mystic flume]

Well, I suppose "free" is relative. If they keep 25% of your royalties that isn't technically free.

[lucid valve]

I'm just waiting for aprooval right now.

[lucid valve]

I dont look for royalties as I dont take these thing seriously, I just wrote the musics and it feels fun to get it out there!

[charred merlin]

Images pushing the link above, but for conversation's sake, Unchained claims to offer 100% on their free account.

But ALWAYS keep this in mind when choosing a service -- CHEAP / GOOD / FAST -- you can only choose 2.

[mystic flume]

All the free services are built to encourage you to pay. Even MusixMatch, the lyrics service, is free... but it took them 3 weeks to "verify" my band's identity, and it didn't go through until I paid the $36 yearly subscription fee. After I paid, it went through the next day. So... that's just an example. You can do a lot of this stuff free, but it'll cost you time.

[lucid valve]

@charred merlin what sould I take screenshot to be safe? Also I have the receipts, does that prove ownership aswell?

[vestal rapids]

I use Soundrop, 99 cents per track released forever, no annual subscription

[lucid valve]

Sorry for the ping!

[vital plaza]

look I paid 22.99 usd, and they did all of the job for me. I hadve 100% revenue from them . believe me distrokid is the right place to go, but if you want a free one, i am not sure i can help you, Sorry😋

[lucid valve]

Yeah I noticed that (MusixMatch thing) aswell on my friends realease

[lucid valve]

It's okay, thank you!

[spice stone]

omG

[spice stone]

thats me !!!

[charred merlin]

No worries! Ping away ❤️ As long as I'm around I'll help out

You'd want a screenshot of the receipts showing the dates and times, and you'd want screenshots of the songs that show the times they were generated (so someone could see that from dates a>b you were covered, and also compare that the song sections were generated during that time on the same account)

FTR, I've had my songs up on DistroKid for a month or more now and haven't had a single peep of an issue, but that's probably really good proactive advice for us to share now that I think about it, just to have any important documents related to your music production tucked away if you intend to release them -- if you all of a sudden get a letter from Youtube that you're being flagged for stolen content and your account is frozen, you might not have the presence of mind to as easily sort back through your library to find everything.

Take 5 minutes today so you don't need to spend 5 hours later (or a stitch, in time saves nine)

[mystic flume]

Then it sounds like RouteNote will work best for you. You can also use BandCamp if you want it out there now. My album is on every streaming service, but BandCamp is still the first google result.

[charred merlin]

Oh! I'm glad you decided to check them out! 🙂

You give up a chunk of the earnings (if you care about selling music), but having the ability to share your music and let people listen without signing into an account is REALLY freeing.

Any feedback on that route?

[lucid valve]

Thank you for the recommandation

[lucid valve]

1 second I will translate this hahahah

[vital plaza]

Nice music! Suno Make you dreams come true 🙂

[charred merlin]

I will show an example ❤️

[lucid valve]

Guys I seem to have an issue... Here's the thing:

My album has 13 musics, and one of them were created by my friend in his account which doesn't have Suno Pro, will that ONE song get me in trouble?

[mystic flume]

I didn't use RouteNote, just looked into it to help answer the question. LANDR does all that stuff for me, even though it's much more expensive than some of the alternatives. Since I subscribed for a year, I've got until next May to decide if that $200 was worth it. Probably not, but I like their AI mastering a lot. Even when I get a track sounding very good to my ears with just Audition, it always sounds so much better after they do their thing. Plus I can jumble all the tracks into a pile and it will master the whole album at once to help normalize things. And the distribution is pretty good, even if I have some issues they've been pretty responsive in emails.

[mystic flume]

Suno owns that song. So I'd leave it off your album.

[charred merlin]

This song would be okay (if receipt was real) ❤️
If they have more questions, I send them that song, and any other parts used to make the song

[lucid valve]

Shit, I already put it in the album and I can't change that cause it's in the distributor

[mystic flume]

Here's what you can do though. Create the song again. Use the same or similar prompt and the same lyrics, and do a bunch of generations until you get one that sounds even better than the first one. This is highly possible because as far as I can tell Suno only creates bangers.

[charred merlin]

EMAIL THE DISTRIBUTOR -- TELL THEM YOU MADE A MISTAKE

IMMEDIATELY so you can prove you fixed the mistake as soon as you learned

[lucid valve]

Okay I will

[charred merlin]

No dudette, I meant BandCamp -- didn't I tell you about them for "first dates, bikes & skates, stay-up-lates"?

[mystic flume]

It's not the end of the world. But that could be what's holding up your release; Content ID checks. I bet if you resubmit the album without that song it'll go through faster.

[mystic flume]

Oh, sure, we talked about it. But I mean... I knew what bandcamp was already. You did convince me to put the music up there, though.

[lucid valve]

Yeah that still doesn't explain cause asesinas is a single (submited earlier than the album it's self) and it still didn't through

And it was made with Suno Pro

[mystic flume]

Well, either way, I'd revise the album and remove that song you don't own. Even if it ends up taking longer.

[charred merlin]

It's a free service, so it's going to take a long time regardless. But the more information they have, the quicker it can usually get processed.

[lucid valve]

Another question, should I send them a ticket explaining that I uploaded a song I dont have the right or should I elaborat that it was made with Suno and the song was made in another account and what not?

[charred merlin]

That's what I meant 🙂 I know that you've really explored a lot of different ways of sharing your music (physical CD, streaming services, digital music stores, etc), and was curious if you had any opinions about the ease of use for BandCamp (besides being annoyed enough at having to manually upload content one by one that a paid membership felt appealing 😛 )

[charred merlin]

I'd say telling them specifically HOW it was made doesn't really matter if it's not legally yours.

Be as informative as possible, relating specifically to "need-to-know" for the matter you're emailing about.

The screenshots and whatnot are helpful to hold onto in case anyone ever asks, but it would be helpful to be direct and only talk about 1 issue at a time unless they bring something else up

ie. I uploaded track 8 {name here} on my recent album without properly obtaining the necessary licensing -- this was a mistake, and I'd like to know if it's possible to simply remove that track from my submitted content, or if it would be better to submit the entire collection again properly this time?

instead of I'm making an album, and I was using Suno to make the songs (I can make music using Fruityloops or playing guitar, even singing in the shower, but it makes it so much easier this way), and I noticed that when I uploaded the song that it was created by a separate account that didnt have the right kind of pro membership, so that song probably wont be allowed, but I hope that i can still keep everything else if thats okay, and i promise that ill pay more attention the next time that i upload everything and also heres some pictures to show that i own the other songs

[lucid valve]

`Hello!

My album (ÉDEN) has 13 songs, and it seems like I uploaded one wrong song.

Track number 12 (adán y eva) was made with an AI account that doesn't have the pro subscription, and it seems I don't have the rights to that song. All other music on the album was made with a pro subscription, so I have the rights to those.

Instead of denying the release, I was wondering if you could just delete track 12 itself. If you need confirmation, the track's name is "adán y eva."

Thank you.`

Here's what I sent, I just really hope I don't get in trouble with Suno itself

[charred merlin]

No, if anything happened regarding Suno in that way, it would likely happen specifically on the platform you're sharing the song (youtube might "talk to Suno" and refuse to host the song, perhaps), but I would imagine the fact that you're taking steps to do the right thing would be a bigger deal ❤️ 🫂

Congrats on releasing your album!

[lucid valve]

Thank you so much, I'm just really scared of legal stuff and what not

[mystic flume]

Well, I wish BandCamp had a MusixMatch integration instead of needing me to add the lyrics manually. I've listed everything as $0.00 but after 200 streams they force people to pay to listen, so I'm not fond of that (though it probably will never become a problem). And yeah, having to pay them $10 so I could upload the entire album at once instead of one song at a time was kinda silly. But I understand they gotta make money somehow since I'm sure 99% of the songs they host never earn any money.

[charred merlin]

I completely understand -- when I released my album, I could hardly think of anything else between when I submitted it and when it appeared on Apple Music -- luckily it was through DistroKid so it only took 1~2 days, but I had "jumped into the deep end" and felt like there were so many new concepts for me to learn, I just hoped that I did the right thing and that nobody would yell at me later 😅

Again, the fact that you're taking the time to think these things over shows that you're thinking the right way -- very excited for you 🙂

[lucid valve]

Hahaha

Thank you so much, have a nice day/night

[vestal rapids]

it's worth noting that Spotify will not pay for a track you until you reach 1000 streams for a track

that can be hard to get

[mystic flume]

This is fine, but they're probably going to tell you to resubmit the album, so if you can create a new one now and save it as a draft or something, that'll save you some time later.

[lucid valve]

Yeah, I'm not doing it for the money and what not it's just fun haha

[charred merlin]

Oooh, for a specific track -- that's an interesting note to point out. Youtube has VERY restrictive rules for their partner platform, though I hadn't really looked too deeply into others.

[vestal rapids]

yeah it's 1000 streams for the track in the last 12 months

Spotify is the biggest platform so it really matters to get those numbers

[charred merlin]

I'm sure that's what most overnight sensations tell themselves -- right up until they have millions of fans waiting for the next cool track to drop 😉

[lucid valve]

Okay, thank you!

[lucid valve]

Hahaha

[mystic flume]

I'm sure the other distribution services do something similar, but I like how LANDR shows me exactly how many streams I've had in this fun little graph:

[lucid valve]

May I see your album?

The names reminded me of my favorite album

[mystic flume]

Sure, link's in my profile.

[lucid valve]

Okay

[shut hollow]

im trying to make a song were instead of music at the end. the guy will " read" from a piece of paper a small sentence.

but im not sure how im suppose to make the prompt

[tight delta]

I haven't used Suno all day, now when I'm trying to generate lyrics with a long prompt, I get this. Any ideas why?

[charred merlin]

https://discord.com/channels/1069381916492562582/1259902708765167709 This is an unresolved bug -- use free chatgpt or bing.com to write the lyrics, or use less than 1249 characters before clicking "generate"

[tight delta]

Ah--thank you!

[charred merlin]

Insanely difficult to "roll" as a "one-shot generation" -- break it down into separate chunks to gain granular control over the song instead of trying to "control" Suno to take specific directions at specific times. You can reasonably "nudge" the direction that a song is leaning by using [short-metatags] in your prompts, but you're not asking for a nudge -- you're asking for a complete change in direction.

Let's break this down to figure out how we'd handle this simple example in 3 parts;

😎 Terminator Extensions

required sections
part 1 - the "intro" -- whatever style of music you want - still referred to as "the intro" for this method even if it lasts 20 minutes long,
part 2 - the "terminator-extension" -- a extension point placed JUST before the timestamp where you'd like to affect a major change -- this will serve as an "interruption point" to prevent part 3 from gaining musical context to build upon, essentially forcing it to treat further extension points in "the chain" to act largely as their own unique generations, without influence from part 1,
part 3 - the "pivot" -- extended from part 2, this will contain the stylistic change from part 1 -- notice that we're inserting a terminator-extension between the 2 desired sections

instructions
1 - Determine what change you want to influence in intro / part 1 (eg. genre swap, forcing an end to the song -- for this example, we'll swap from opera to metal halfway through)
2 - Generate or select your desired intro / part 1 (eg. opera song, 4m)
3 - Create an extension point, part 2 just before you'd like to affect the change in intro / part 1 (eg. we want to swap halfway, so I'll place the terminator-extension/part 2 at 1:59, or perhaps a few seconds earlier if we need to revisit this step and refine for better results)
4 - Create an additional extension point, the pivot / part 3, timed at 0:01 of part 2 -- this only gives the pivot / part 3 a single second of musical context to "guess" how the rest of the song should sound -- the song's momentum is essentially at a standstill, and you have remarkable power to decide where it goes.
5 - "Get Whole Song" from part 3

provided example

intro = 0 > 0:31
terminator = 0:31 > 0:32
pivot = 0:32 > 1:32

[shut hollow]

ahh thanks. I can see the idea in it thou. Time to be testing 🙂 Thanks!

[stiff sable]

just a quick insight from me: if you google about routenote and AI music there are quite a few reddit posts and online threads about routenote rejecting AI generated music outright which could be potentially part of it. whether its an automated process or some staff are listening and reject it whilst others are accepting it is hard to know. i dont know that for sure but im going off of what i read, which is fair for a free distributor and less-than-clear AI legislation atm. :)

[lucid valve]

Ohhhh okay

This actually gets me scared LOL

[charred merlin]

You're welcome! It might take a little wiggle room to get it working the exact way you want between implementations, but the basic idea still applies.

If you notice the termination point between the "intro" and "pivot", it has a smooth(?) transition between the two, and if further predictions might save you any wasted time or credits, I might suggest trying to make sure that the "pivot" starts from absolute silence, so that any noise isn't extrapolated upon and diffused into more instrumentals -- unless that creates a cool creepy ambient vibe in the background...

There's no "guaranteed way" to accomplish even this, and I urge people to practice with it near the end of their daily/monthly credit cycle as a means of experimenting, familiarizing yourself with new approaches to problems, and building your own mental "English-to-Suno" translation guide -- the key steps to being a Sunovator.

Either way, I wanted to note that attempting such a strong "change of direction" as going from music to spoken word narration would probably be handled by "sending a second terminator in after the first one..."
-# ...which is the exact plot of Terminator 2 : Judgment Day, now that I've said it out loud and thought about it....

[stiff sable]

@lucid valve its nothing to do with you specifically dw! they arent going to punish you for uploading anything; just letting you know some other ppl have had issues with routenote so you may just get an email saying your content wasnt approved 🙌 you might be lucky tho so keep the upload up and see what happens!

[lucid valve]

Don't worry! I'm scared they denied one of friend's album

But it's weird cause his first release got accepted and it's now on spotify and stuff

But the second didn't

Both made with AI

[stiff sable]

ohh thats curious! that definitely leads me more to think that its something to do with a human verification process and some of their staff are accepting it as long as everything else is fine whereas others are just rejecting AI content completely

ive tried routenote before for my own non-AI music and they do just take forever bcos theyre one of the only distributors i've found where they still manually human-review every single track before accepting it

[shut hollow]

yea. im just gonna start with like last chorus and build inwards in the song. i got enough points to play around for a bit atleast 🙂 again thanks alot! i was having a hard time working it out my self.

[lucid valve]

Oh wait they check it manually????? Holyy that really is outdated

Yeah that makes way more sense cause my friend's album (both accepted and rejected) took forever

What's sad is that they take forever even to review even if it rejects the release... I really wish I could just cancel my release with them so I can upload in another one...

[stiff sable]

yeahh i dont think they listen to the whole song as that would be pretty ridiculous... from my understanding by emailing their support team they do all the copyright checks automatically etc, its more of a manual metadata check but it seems they do check a portion of the song to ensure it is actually musical content and not just anything which is maybe where the AI generations are getting caught

[lucid valve]

I really hope that they send my result sooner cause I have a sales date in mind :((

[stiff sable]

yeahh this bit is rlly annoying i completely get your pain with this. waiting all that time just to get a rejection email is extremely annoying

[lucid valve]

I'm really considering buying it DistroKid

But I'm afraid of getting 2 albums that are the same in my spotify profile

Idk if that's possible

[charred merlin]

You can link existing content to them

[stiff sable]

i havent tried them yet but i have heard good things about Jumpstr (which is also free). I made an account with them and took a look around their upload tools so that could potentially be an alternative to look for? If you're uploading lots and doing it seriously (like 1+ release per month) then Distrokid is defo probably your way to go.

[lucid valve]

Hey does DistroKid have a monthly plan?

[stiff sable]

only thing i dont like about distrokid is their interface lol, i much prefer LANDR's or United Masters (which i currently use)

[lucid valve]

I don't really wanna pay for a yearly plan

[charred merlin]

https://distrokid.com/plan/ sorry -- nope -- cheapest plan is $1.92/month billed 1 year at a time. You save 7% if you're referred by an existing DistroKid member, and I think they get $10 out of it, so maybe you can work something out with someone?

[stiff sable]

but if you only need the one artist name to release under, then the cheapest basic plan will do what you need - don't let the more expensive ones overwhelm you just bcos they offer them

[charred merlin]

I went with the middle plan because it came with a custom label name, so I put my wife's small business name on there and told her that she's now also the owner of a small indie record label.

[stiff sable]

omgg i love that 💗

[lucid valve]

Okay thank you!

@charred merlin sorry for the ping, do you recommend any others distruitors?

[charred merlin]

I really only have familiarity with DistroKid, as they're my distributor -- not trying to shill for them, I made comparisons on a few competitors based on what I wanted from my distribution partner.

I found that DistroKid is very fast (Most services within a couple days, large remainder within 1-2 weeks), very affordable for the services provided (one time yearly fee, then zero hassle, a bunch of perks), and they have a very modular set up (meaning you get to choose which extras you want, instead of paying for things you might not need)

[amber vigil]

I’m pretty unsophisticated on actually making quality music for distribution. Would you mind explaining what you mean by AI mastering? And why do you find this important?

I’m thinking of polishing and distributing some of my songs as well. I appreciate you all for sharing your insights on this!

[mystic flume]

Well, I guess the easiest way to describe it would be to show you. So, after i've finished editing the song and making it sound as good as I can in Audition, I put that result into LANDR's AI "mastering" engine which makes it sound even better.

Before:

After:

I can't say for sure it's a result I couldn't achieve in Audition, but I'm a novice at music mastering so having that extra bit of "expert" tweaking feels like it helps.

[charred merlin]

Upload one of your songs here, and choose from one of the 4 free auto-mastering mixes to see which you prefer -- or you might prefer the original.

AI mastering, or auto-mastering, is like when a band finishes recording their individual instruments in the studio -- now it's time for the highly paid audio engineer to adjust a bunch of audio levels, typically referred to as "mixing and mastering".

Since each engineer eventually becomes famous for their own particular areas they accentuate in a song (like which instruments stand out, eg. listen to how different Metallica's "...And Justice For All" sounds WITH bass, or how to direct a particular live recorded performance in addition to the mixing that follows, eg. the John Peel sessions with The Smiths)

Important thing is not to overthink it, as there's no such thing as perfect -- as long as the song sounds subjectively good to you, it's "good music".

Plus, if you release it now instead of overthinking it, you get to return to any mistakes that end up haunting you and release the "REMASTERED" version of your album -- which sounds pretty dang cool.

[rustic prairie]

has anyone found random artifacts at the end of thier songs. I have a looping nonsensical rant at the end of this: https://suno.com/song/520c52dc-4822-4be5-99c6-e9f2ef633bff

[charred merlin]

lmao, this is one of the longest, most intense hallucinations I've ever heard Suno make.

[rustic prairie]

I know right, it almost veers into english, but it's nuts.

[charred merlin]

... > Report > Bugs > Doesn't Follow Lyrics, etc. as a means of trying to prevent hallucinations like that in future models (but where' the fun in that?)

[rustic prairie]

it's actually amazing, so far off the prompt

[charred merlin]

There was someone a few days ago arguing that Suno was hallucinating because it repeated a verse instead of following instructions, while almost every other requirement was met -- they didn't quite agree with my viewpoint that it was "a minor hallucination", and pressed the fact that there are no such distinctions.

I'd argue this counts as a moderate to major hallucination 🤣

[charred merlin]

https://tenor.com/view/anchor-man-ron-burgundy-ron-not-mad-amazing-gif-8937476

[rustic prairie]

agree. The song also went way in a different dirrection then what I expected, in a good way. We need to look for patterns in the hallucinations

[charred merlin]

That's a big ask.

Earlier today I was staunchly arguing that the models aren't fiddled with, but even I'm seeing some wild stuff -- a lot of changes from day to day, things that didn't act weird are starting to act weird, even for me.

[unique stone]

I love hallucinations. I deliberately try to force them sometimes (instrumental, female vocals...)

[rustic prairie]

I've made a LOT of songs and have not encountered this. I'm going to try to force hallucinations now

[unique stone]

You can also try extending from the middle of the hallucination to see what that leads. I had some kind of nonsense poem being read once

[charred merlin]

I don't want to ask you to burn credits trying to recreate it, so I'm trying a few generations to see if I get anything weird.

[naive quail]

grumble

[charred merlin]

...wtf is going on? I have no idea how to explain this, and I've never heard this kind of noise before -- even when overloading and intentionally trying to break Suno (while abiding by ToS)

https://suno.com/song/5b9ee576-92b4-4293-8b1d-a2a576ae3c6b

Does anyone else get a weird glitched noise ~0:10?

[naive quail]

I know this sound! that's the cat f'ing with the mic while you're recording.

[golden plover]

Hi guys there’s a track I made in suno, I really like the drum kit used. Is there a possible way someone can decipher what drum kit was used? This is something I would like to use for one of my own songs.

[charred merlin]

https://tenor.com/view/skeletor-masters-of-the-universe-ha-missed-you-missed-missed-gif-18421113

[charred merlin]

Hi! While I don't doubt that SOMEONE here might know, tech-chat is more directed toward questions regarding technical-support and best-practices for using Suno -- you might have better luck posting your request in #💬┃general-chat #🗣️┃chill-chat or https://discord.com/channels/1069381916492562582/1154862381407666276 ❤️ 🫡 🥁

[golden plover]

Ok I will repost there, thanks

[charred merlin]

@unique stone , @teal raft can you try something for me?

[stiff sable]

ohhh boy the other day i generated this song, it was one of the most beautiful things i'd made with suno, let alone heard ever; so it gets near the end of the song, then suddenly the vocals stop and the music cuts out, there's an almighty GASP and then SNORING and then the instruments start up again before fading out 🤣 idk what happened to the singer but they were out cold!

[lethal ore]

Roland TR-808 and or TR-606, sampled hats

[charred merlin]

@hollow hare also pinging to see if you're around to check something

[rustic prairie]

I hear it. Maybe context, maybe not. I was using a hack I discovered to request it sound like a specific artist. Don't think that's it, but may be useful.

[charred merlin]

Hell, @untold briar, @wet minnow (only pinging mods I see listed as online at the moment), @flint haven , and anyone else who has ~30+ credits to spend on an experiment?

[hollow hare]

I have 20 credits to spare until renewal in 2 hrs

[untold briar]

Sure, but a little later tonight

[wet minnow]

yes

[charred merlin]

No worries -- I'll post once while you're here, the ask is VERY simple when you get a chance later -- I just need to you spam-generate underprompted v3.5 songs -- based on my observations, v3.5 underperforms/produces "samey" results when used without very effective use of prompting, or a strong enough melody to "hook onto" -- as I've been saying since it came out.

[untold briar]

That fits with my general obversation about converting V3 prompts to V3.5 (make them stronger)

[wet minnow]

what? 😄

[charred merlin]

I would really appreciate it if you could share the results here after you try. If you want to follow the exact testing methods I'm using, simply add "test" in both the lyrics and style prompt.

Specifically, spam like 3+ creations at once with this incredibly undereffective prompt and let me know if your results look like this -- don't mind the weird titling, I wanted to make a observational notes from day to day regarding how this issue evolves, for my own testing purposes.

do you see anything weird looking through them?

[wet minnow]

just the same as we discussed in "tonys lounge"

[charred merlin]

I completely agree -- v3.5 is an improvement in that sense, similar to dalle3 in that it requires effective use of prompting, otherwise it underperformed?

[wet minnow]

i have the same results

[charred merlin]

Could you share for posterity? ❤️

❤️

[hollow hare]

To clarify, you want us to use a one atypical word (at most? Exactly one word? Fewer than three?) style and lyric prompt to see their outcome in 3.5?

[charred merlin]

@hollow hare, if you have different observational data to include, I'm happy to compile it all 🙂 Even if it's results that show otherwise, or any ideas to build upon to critique my methods -- I'm happy for any collaborative efforts that you're willing to engage in

[wet minnow]

also have a look at the room where we discussed this. we made experiments there too

[wet minnow]

i will ping you from there

[stiff sable]

can i partake in this out of pure curiosity 🤧

[rustic prairie]

I clearly need to hang out here more. I've had amzing results from both complex and simple prompts. Need to understand.

[charred merlin]

Sorry, I misunderstood the situation and thought there was more research to be done, but I think I'm overstepping.

[wet minnow]

i didnt wanted to stop you from doing additional research. just show you the results of a similar test that basically confirm what you also see.

[charred merlin]

It's okay dude, you guys have it -- just have them let me know if and when they need help on the matter, until then I don't want to step on toes ❤️

Apologies for any weird vibes -- just overzealous, evangelistic about Suno, and wanting to apply myself as best I can, and seem to be spending a lot of energy in the wrong ways.

[wet minnow]

your test confirmed what we suspected. so it isnt wasted. Because there is no official way to tackle problems i also often do things that feel like wasted time and effort. but it isnt really that. at least i learn from the things i do. and sometimes i have to do them by myself to learn something from them.

big hug.

[charred merlin]

Hey @I am Blue -- I am also blue!

Luckily I posted a guide on that exact method earlier today! I like to try to coin new terms, so I'm referring to it as a "terminator-extension", primarily because it was initially used to force the end of a song, but it's also really helpful for making songs turn-on-a-dime -- ⁠#📚┃suno-school message

In your case, for part 3, or the pivot, you'd want to strip all lyrics, and only use a single word style prompt with v3 (as v3 typically reacts better to this method) -- I'd suggest something like "end", "goodbye", "so long", "farewell" "auf weidersehen" "adieu", etc

[wet minnow]

i think the link to you workflow guide broke while editing https://discord.com/channels/1069381916492562582/1256301881966198805

[charred merlin]

Hi Pirolyx, I think the issue is at least on the developer's radar, but at the moment all I can suggest is that we continue to experiment.

If it counts for anything, I tried testing the ONE prompt that I'm hyperfamiliar with, and I'm getting an average of 1m56s for a v3.5 generation -- I'm going to look back through previous times I've used just this prompt and see if I can compare or contrast anything -- and then I'll look into it periodally to figure out if there's any form of persistent or emerging issue.

I know you've been dealing with this issue for a bit, sorry I wasn't able to look into it sooner, my friend ❤️

[viral lotus]

@charred merlin if you have a moment or two, could you help me diagnose a problem I'm having with a song generation. I think I know what the answer is, but your opinion would be immensely helpful.
https://suno.com/song/4bc40bd6-4d9c-44fb-8d7a-5d8d246f4c9c The first problem is in the transition from Pre-Chorus to Chorus; and the second problem is that the Chorus is not generated like a Chorus. It sounds just like a verse. (The chorus is relatively the same in every generation)

[charred merlin]

Hrm, unless I'm missing something at a certain timestamp, I don't notice anything particularly "off" about it?

Could you elaborate a bit more on what specifically is "grabbing your ear", and what you were hoping to achieve instead?

[viral lotus]

First, there is a huge pause between pre-chorus and chorus. Second, the chorus doesn't change much at all compared to verse or pre-chorus. Same vocal range, pretty much the same pattern, then it dips lower.

the pause happens every generation, as do the rest of the "problems"

[charred merlin]

Just following the cymbals, they're sparingly used until the pre-chorus, when they slowly fade in, and are a constant presence in the background of the chorus -- which seems to have a nice dramatic pause to emphasize that it's a separate measure, intended to convey a different "scene" if you will, but the lyrics continue on.

I suppose it's a subjective matter, but I genuinely think it sounds nice; "underneath these ---", even by itself would make grammatical sense as if you're singing about the lights to the lights.

Where are you starting these generations from -- sorry, I can see the extensions in the video ❤️

[viral lotus]

I like a pause, but not a 30-second pause (OK, little exaggeration).

***and don't mind of course the duplicate verse

I did this before I went to bed a couple nights ago so don't really remember specifics of generations right now. I do know I stayed up three hours late doing something else, then had this idea for this song and stayed up at least another hour

[charred merlin]

Oh my god -- I must be tired -- hahaha, I'm sorry dude

I just realized I was paying attention to the transition from verse 2 into prechorus, thinking "...dude, this is perfect. Maybe there's a big dramatic pause, but what are you complaining about with this first thing?"

Okay, everything makes a lot more sense now 😛

[viral lotus]

I like the pre-chorus fine. OK, I'll give you the pause itself is not horrible, but listening to it again...I guess the real problem is the pitches of "nashville lights". Don't you expect it to soar a little?

yeah, I like the prechorus OK, no real complaints

the chorus is just flat

[charred merlin]

hahaha, I totally misunderstood that, my bad -- okay, I have an idea or two, let me see what works better

[viral lotus]

sweet thanks

[charred merlin]

Foggy! ❤️

[viral lotus]

I need to start learning how to embed metatags in the lyrics. Have no idea what I'm doing right now, just throwing stuff against the wall

[amber vigil]

i’m unaware what Audition is. But i think i have the gist … you make the song in Suno, use Audtion to polish it up a bit (?), and then use LANDR’s AI engine to make it sound a bit smoother. Is that correct?

Could you skip Audition and just use LANDR?

[amber vigil]

i think i need to get my better headphones. It’s really hard to hear in sufficient depth with my travel headphones. I appreciate the detailed explanation though and will most likelymlearn more as i try things out.

[charred merlin]

No "real" trick to it, they're just "localized" style prompts.

If you want to change the style at a specific part without effecting earlier sections (...okay, maybe effecting earlier sections?), you can add anything you would add as a style prompt into square brackets, and [supportive] it does everything that it can to help and try to embrace the following lyrics lovingly

There's tons of experimentations with metatags, regarding what does and doesn't work (unfortunately trying to compile a list of such is like asking "what does and doesn't work on chatgpt?")

https://www.suno.wiki/faq/metatags/ there IS the community wiki, but it's by NO means a comprehensive or exhaustive list, and should be used as a jumping-off point for creativity 🙂

https://tenor.com/view/let-me-be-clear-bart-school-say-the-line-bart-say-the-line-gif-4033914080036998615

..."let me be clear -- prompts are suggestions, not commands."

[viral lotus]

yes thanks, read all that...I ran into a rather large list of metatags about a month ago, many of which I hadn't seen at all/a bit unusual. But my computer reset and I lost the page it was on. I searched and searched, but couldn't find it. I was fairly sure it was reddit, but spent a while there and didn't find it.

[amber vigil]

freaky. I had a super odd one where, at the end of a very depressing song, the AI exhales and gives an odd laugh and then a sigh. I left it in because it sounded cool as if the singer was releasing the tension found in the song.

[viral lotus]

and yeah, I understand "prompts not commands". I've spent more time using chatgpt than anything in my free time over the last year.

[charred merlin]

...yeah, there's a lot of people who believe they've "figured it out" by "coding" Suno. I'm not even saying it at anyone in particular at this point, I just feel like it's a public service.

[viral lotus]

and Suno is a bit more "creative" and follows directions (as we understand them) a bit less

I just want to go through a bunch of songs and collect metatags and see what their effects are.

[charred merlin]

Every time I DON'T say it, a green-lead discorder shows up, looks at a trending song that says override.ai == make song awesome ; <emotions : sad and angry -- 60%> wubbing rittittittitti noise that increases in static until nosebleeds in the metatags and they think "OMG, that's why all of these songs sound great and I'm not getting perfect results on a one-shot generation, it's because no one told me the secret!"

[viral lotus]

And then cram a bunch of them sequentially into songs to experiment.

I didn't understand a word of that, but you are joking?

oh, yeah, i got the sarcasm in the last line now, not the rest, lol

OK, I'm slow...got it now

[charred merlin]

No. People legitimately put that and much more ridiculous stuff into it.

And people legitimately believe that the trick to making "a good song" isn't experimentation between generations, collaborating, breaking your requests down into smaller pieces to understand what changes you're actually tangibly making -- or whether you're just subject to confirmation bias like every other human on earth, especially when emotionally invested.

(again, not at anyone in particular.)

Just putting it out there, if someone tells you "they learned secret prompting tricks from the guides/devs/mods", they're straight up lying to make themselves sound "exclusive and cool."

[amber vigil]

howdy! Checking in on the techy stuff … building my skillzzzzz (hopefully).

I learn so much in here

[viral lotus]

Yes, of course, the way it is "programmed" (trained) it is a black box

[charred merlin]

Even the devs are learning some stuff as they go, and there's a lot of collaborative learning.

When people tell you that they "completely understand" something in Suno, ask them for a simple, easily demonstratable "prompt snippet" you can test on it's own to see how it works (maybe at the end of your daily or monthly credit cycle as not to "waste" your leftover credits).

If they're not able to provide you with a "snippet", they likely are focusing on the 1 thing that worked, instead of wondering why everything else they prompted was ignored.

[teal raft]

The devs that actually build Suno:

Maybe?

And I should know, I am exclusive and cool. Like harber

[charred merlin]

😛 It wasn't you, but that time it WAS intended to dispel a certain rumour that someone was spreading to gain clout.

Sharing fun ideas and trying new things to see what wild results you can get out of Suno is one of the most amazing things about the platform, and if people are inadvertently misdirected away from reapproaching the situation from an open-minded and experimental fashion, it's not maliciously done.

HOWEVER, actively spreading misinformation, or outright fabricating incidents in an attempt to lend credence to your ideas, that's actively doing harm to people who genuinely want to learn.

[charred merlin]

[viral lotus]

Did you see this one: https://suno.com/song/945b8e58-5228-4687-824b-b8addf10ad83

I tried to make it as horrible as possible, but it's still interesting lol

[elfin storm]

I want to ask if the lyrics are read by the vocals, how should the lyrics be sung by the vocals?

[charred merlin]

I'm barely able to form coherent sentences, so I think it's time for me to shove Woodhouse off the couch and grab a couple hours of sleep.

I'd love to pick your brain regarding your take on LANDR, I'm taking steps to learn more about it -- understandably from a personal perspective, I've been looking at Suno through the eyes of a singer/songwriter reemerging into the music industry -- previously I'd looked at it as a strength that I had experience, but there were some interesting conversations today that made me realize it may have been putting blinders on myself by focusing on that aspect and how people use Suno instead of why they use Suno.

Understanding what roles different distribution and creative platforms like BandLab and LANDR play in the minds of different types of users (that lurker, user, superuser, stakeholder thing I ramble about) could really helpful in trying to make sure that I don't just give someone the advice "that's best for me".

Either way, I apologize that I'm darting off again, but I'm looking forward to the (presumably hours-long 🤣) conversation ahead one day soon 🙂

[charred merlin]

I LOVE using "experimental" as the only style prompt -- sometimes I'll play "Suno Roulette" and just type a train of thought ramble into it, and generate a handful of results -- it's so much fun having NO idea what will come out.

So friggen cool when you realize that you can strip out vocals or something, clip out a cool part, and feed it back into Suno to make a legit-song out of a weird meme track 😄

[charred merlin]

By default, anything that you place into the "Lyrics" section of the "custom mode" will be used as vocals

[mystic flume]

You could, but the quality of the WAV file you get from Suno is kinda bad. So using something like Audition or Audacity to add compression, etc, will always produce a better result.

[elfin storm]

But I always make it to the middle of the verse and the opening is really fast

[mystic flume]

Bear in mind, LANDR is not cheap; there are other options. I'm not over here recommending anyone use it. It's just what I found first that worked for me.

They just introduced an opt-in "let us use your music to train our AI model and earn money!" thing which is... interesting. I can't imagine it'll make anyone money but them, but I'm kinda pessimistic about these things (and the universe in general when it comes to money)

[charred merlin]

@mystic flume, makes me happy to see your name pop up in chat before I scoot off for the night 🙂 -- I would hate to impose on your endless patience for my nonsense, but you're the resident LANDR expert in my mind, I would really appreciate if I could pick your brain regarding it at some point tomorrow or in the near future?

I have a great deal of professional admiration for how you represent yourself musically, and as such your opinion would be exceptionally meaningful. 💜

[mystic flume]

Well it's true. I am pretty great.

[teal raft]

I tell everyone the same thing about myself and they usually say "who are you?" and "go away"

[elfin storm]

This is what I mean
https://suno.com/song/07238865-b485-4b68-98c1-2b76d4fc7466

[charred merlin]

and "why are you standing over my bed like that, Kaker" and "what time is it"

[mystic flume]

jokes aside, you know where to find me

[charred merlin]

Definitely something I'll be bringing up, but from what I gather at a glance, they're calling it "Fair Trade AI"? ☕

You retain full-ownership, you have flexibility in the way you're able to out-out, and it's supposed to enable revenue sharing for musicians (still trying to wrap my head around how making a ledger of things with blockchain makes payments easier, but I'll get there someday, probably).

Okay -- thanks Tinda, you rock 🙂

[charred merlin]

It seems like including the words "mid-tempo" in the style prompts might be causing the song to speed up? Perhaps Suno is misinterpreting this, not as "an instruction" to make the song mid-tempo, but seeing the word tempo makes it think about other songs where "tempo" might be how you would describe the song.

For example, if you start dancing at a club with friends, and the song gets TOO fast to dance along with, if someone asks "Muji, why aren't you dancing?", you might say "TEMPO!" -- in the same way, when you put "tempo" in the style, maybe Suno is thinking "AH, yes -- they talk about "tempo" for songs that go very fast, that much be what they want".

Suno is mistaken, but instead of telling Suno that it is wrong, we need to "forget" that word when trying to talk about the speed of a song.

Or maybe we need MORE style words to go beside "mid-tempo" so that "tempo" is only a small part of what we are asking.

[lapis quiver]

Can anyone tell me why my website won't load, I'm in China but I have my VPN turned on

@lapis quiver

[hollow hare]

I assume you got a 4xx or 5xx error from Cloudflare? It usually is a DNS error that is resolved by refreshing in a few minutes (worst case is 10-15 min)

[lapis quiver]

well. Thanks a lot, let me have a try

[amber vigil]

my real question is if landr can do what audition does? I’m guessing by the answer … no.

[mystic flume]

I'd suggest grabbing Audacity, which is basically the free version of Adobe Audition, and learning what sort of things audio editing software can do for you, before deciding if paying for something like LANDR is necessary.

[inner socket]

I have a serious problem. When I compose with suno, there is a high probability that the generated song will be less than 60s, and it appears to be a lot more when using the v3.5 model, e.g. https://suno.com/song/b232ab47-a215-4ef4-a6f9-79f35216cd0b, which is a pathetic 25s. I can be sure that the fill-in lyrics, title and style words are all formatted correctly and are all the right length, so why is this a problem????

[hollow hare]

You cannot get copyright from Suno, that is something you get by registering with your country's intellectual property office.

In order to register for copyright (and monetization in general), the song must be created during a subscription period. Any songs made during that time is yours forever.

Songs made during the free plan are Suno's, not yours, so you cannot monetize the ones made during the free plan.

[foggy wing]

I believe Suno writes that songs created on a free account will be copyrighted to the account holder if the account turns into a subscribed one.

[hollow hare]

*Commercial rights and not according to the FAQs. Did they say that in an updated TOS? If so, please link me to the section and paragraph so I can refer to other people in the future.

You could contact [e-mail] and ask them directly. However, if you do not have it in writing from Suno that the free song can be monetized/treated as if it was created during thr subscription period, this will supercede that

[mint crag]

generate worked for me a few times and thens stopped when i try on the web iphone or rabbit r1 it just hangs and does nothing or I get a 403 error on iphone web and rabbit i have a pro subscription with 2500 credits and I have tried liek 50 times but cant do anything

[foggy wing]

This is very interesting. It has undoubtedly changed the last few weeks. I carefully read the agreement and bought the subscription solely so I could download the wave-files of my free made songs to export to other services (which has not been done yet).

[hollow hare]

That entry of the FAQ has been there for, literally, months at the very least. (E.g. see my screenshot posted at 2024-04-12 at #💬┃general-chat message ).

I took a quick glance at the TOS and found no mentions of it. This was the closest I found, and again, it was a quick search

[foggy wing]

I belive I read it when signing up for the subscription, I will reasearch more.

[hollow hare]

The best source of confirmation would be to contact [e-mail] if this is your case scenario. If it is true what you said, you will have it in writing from the official email

[foggy wing]

The only email I got when I subscribed was the invoice.

[foggy wing]

Also I see no way of actually viewing when a song was created, neither with which plan it was made in or which date it was made.

[upper arrow]

https://suno.com/song/200f8404-5b78-420c-bdf8-bbc1d1dc36bc

I created new lyrics for this song, but it seems like Suno created an all new melody. I wanted to preserve the tempo and melody. What to do?

[autumn jolt]

This is subjective, but after spending a while putting it through its paces, v3.5 is a big improvement over v3 in my view. Biggest change: far less distortion and artefacts in the backs of tracks. That's the second biggest issue I've had with Suno so far. If they improve vocal clarity with v4 my head might explode! 🤯

[foggy wing]

My take on vocal clarity is that it largely depends on what other prompts you distort it with. If you are too detailed it will make the vocals worse, there has to be some wiggle room for the randomness for Suno to generate good vocals.

[mystic flume]

Best way to keep the melody intact is to extend from some place in the song after the things you want to keep have been established, then just start over with the lyrics from that point. Try to keep it within the first 2 minutes, since that's as far back as Suno will look. I've had a lot of success this way. It may take a lot of credits to get exactly what you want, though, so persevere.

[mystic flume]

The TOS regarding ownership of Free songs hasn't changed since at least May, which is when I signed up.

[autumn jolt]

Yeah, I do kind of agree. I've been using the tag [Chorus, Break] recently to get softly sung choruses which then build up again. Without as much music in the background the vocals are definitely clearer.

[upper arrow]

I tried... but they always stuck at 2:00 duration

[foggy wing]

Weird, I wonder were I got it from. I am absolutely certain I read it, since I've been using the free version for a couple of months building several albums to then go over to the paid.

[unique stone]

Ownership of free/subscription songs has not changed since at least November last year when I joined.

[mystic flume]

yeah, that's as much as you can get from an extend. Then you can extend from the end of that one with whatever got cut off.

[foggy wing]

Guess I'll just have to remake the songs made in free mode, but that is hard work since there are no timestamps on them.

[mystic flume]

So, you made all your songs for free, and then hoped to gain commercial rights and download the WAV files of them all after subscribing for one month? Am I understanding correctly?

[foggy wing]

Not really, I started with free, then bought a yearly recurring subscription, I plan to continue with it.

I just did not know how far I would have taken it until I realized I made songs daily using the free version.

[mystic flume]

The good news is, as long as you're able to forget the original versions, you can probably make even better versions using those lyrics now that you have a bunch of credits to use.

[foggy wing]

The credits is not really the issue. Free plan is 1500 credits monthly, the mid paid plan is 2500 credits. Not that big of a difference.

[mystic flume]

Well, the big difference is not having to wait 24 hours to recharge your free credits every day. I probably spend about 500 - 1000 credits per song, for instance.

[foggy wing]

Yup, I recognize that.

[mystic flume]

Good luck with the new versions. 🙂

[foggy wing]

Thanks. If only I could see which ones are made using the free plan.

[mystic flume]

Timestamp is in the details when you click on a song.

[wet minnow]

*But do that on the first part of the song, thats the creation date. The whole song will show the date when the whole song was made. (thats why we often say "A free song cant be turned into a paid song by just extending it")

[charred merlin]

Can’t it? I’d like to be certain before advising in an open and shut manner.

As I currently understand it, since Suno owns the free song, extending from it should function the same as extending from a song that a separate user shares.

The only “definitive” answer I’ve come across is that you should consult a legal expert if you want a clear cut answer as to what degree of “ownership” you have over a song.

At least at this point, I haven’t received any form of clear answers on the matter, so I’m not comfortable acting as an expert on the subject.

I know @unique stone has mentioned that in at least once instance, Suno has negotiated rights to songs made on Free accounts, but anything beyond that would be speculation, and wouldn’t really serve what I’m hoping to accomplish with my support methodology of being a trusted “non-expert”.

[foggy wing]

A side note (pun intended). Has anyone gotten success with key and scale prompts? For example, I find that "[b-minor], [aeolian]" in the Style-box gives different results than when put in the lyrics box. Also, would Suno accept "[b-minor, aeolian]" as a combined prompt?

[naive quail]

If I do enough generations, I eventually get one that matches.

[charred merlin]

It absolutely would, but keep in mind that neither prompt is a “command”, but suggestions.

The important thing to note is that it’s different from the original prompts you’re trying.

If you’re not familiar with “tokenizers, they can be really helpful in understanding how AI interprets things differently than we’d expect

[foggy wing]

Thanks, where can I read more on tokenizers?

[charred merlin]

😛 darn it -- you sniped me in the seconds between my mobile post and desktop edit

https://tenor.com/view/alf-hes-quick-ill-give-him-that-hes-fast-he-is-quick-gif-26039736

#📚┃suno-school message This is from before I went back and tried to "tidy things up", so let me know if it's not a super helpful example and I'll try to make a more comprehensive version and link to it in #1256301881966198805

[foggy wing]

Much abliged. I will read into it.
Currently trying to get Suno to interlace swedish folk melodies on fiddle into melodic death metal, but usually it mixes so the fiddle plays the guitar parts and vice versa. Unique sound, of course, but not what I'm going for.

Entering the scale in the lyrics box made the whole song go full ancient egyptian.

[naive quail]

To Suno that's probably just a bunch of random letters/symbols, which basically becomes "prompt noise". It'll try to do SOMETHING with it which can often turn into something unique and interesting, just rarely what we actually intended.

[foggy wing]

Yeah, more than once I've had a chord sequence turned into vocals. "Eeeee maaaaajoooor, seeee shaaaarp". 😁

[naive quail]

lol! It's hilarious when Suno vocalizes the letters/symbols that make up ascii art. This is Suno singing all of the W's that make up the pot leaf https://suno.com/song/013cd611-e476-47fe-8d6e-d1afec929ee8

[charred merlin]

This happens if you overload metatags, but I've seen a LOT of recent instances of style prompts bleeding through to the vocal hallucinations.

I'm to the point where I'm not trying to figure it out anymore, and just pulling up a chair at our tech-support-therapy circle for anyone else trying to make sense of things 😛

[naive quail]

Goats on Welfare!

(text in the Style field that Suno actually sang)

[charred merlin]

https://suno.com/song/f735599e-cdf9-49dd-aa21-984fe7a9cdbb Just because it's a banger that never dies, and should have handled the "can suno read emoji" conversation right in it's tracks

[naive quail]

screams

[charred merlin]

lmao, of course you added a Kilroy

[naive quail]

Not sure what Suno made of that 😄

[charred merlin]

when the Kilroy is pronounced as "Six String Toker was here"

[naive quail]

hah

[foggy wing]

Continuing the scale, key, tempo, rythm discussion...
I've noticed clear improvement on getting the desired outcome when slowing things down with increased resolution. 4/4 120 bpm will sound really messy and flat and can make slow instruments like strings behave weird, but do 16/16 60 bpm and it will have pretty much the same feeling but way clearer sound image.

64/64 35 bpm can do some truly fantastic things.

[long viper]

oops, wrong room

lol

[inner socket]

I have a serious problem. When I compose with suno, there is a high probability that the generated song will be less than 60s, and it appears to be a lot more when using the v3.5 model, e.g. https://suno.com/song/b232ab47-a215-4ef4-a6f9-79f35216cd0b, which is a pathetic 25s. I can be sure that the fill-in lyrics, title and style words are all formatted correctly and are all the right length, so why is this a problem????🤕

[foggy wing]

Try adding tempo, see what 60, 90, 120, 150 bpm do for it.

I format mine like this on the first line of the lyrics box:
[key: B-minor], [scale: Aeolian], [rythm: 48/64], [tempo: 35 bpm]

[unique stone]

Suno completely ignores tempo. The numbers mean nothing to it. Typically you'll get a tempo that is "appropriate" for the genre.

[naive quail]

Adding any tags before that first verse might help, such as [Instrumental Intro]. I'd also try changing that first verse tag to [Verse 1]

[unique stone]

Only way to guarantee a particular tempo is to upload a metronome sound and extend from that

[naive quail]

One weird thing that I keep noticing is that I can never reproduce the short track issue when using the same exact prompt.

[foggy wing]

"completely ignores" seems a bit exaggerated. I concur it may not be spot on, but if I generate 4/4 120 bpm and 4/4 85 BPM I will get one faster and one slower.

[charred merlin]

There's some ongoing discussions, and all I can say is that I think it's on the dev's radar, but I have zero context -- before yesterday I was a staunch advocate that the model didn't change from day to day, because there were daily reports of people having an underperforming result, but failing to properly explain how they achieved that result.

Last night I noticed that the only prompt I'm comfortable claiming any form of expertise regarding is only giving me ~2m generations on v3.5, whereas I was getting ~>3m30s generations before.

While I can be candid at times, it's very important to me that my advice is viewed objectively and not simply as "theharber.'s opinion", so I hope that people constructively criticize and build upon my methodologies so we can all benefit.

[charred merlin]

From everything that @unique stone and I have thoroughly tested over the past months, it's using the word "tempo" to suggest that it needs to do something related to speed, but please understand that Suno doesn't accepet "commands" or "codes" in this way.

We encourage you to do your own testing and make observations, but if it helps, we've tried many ways to accomplishing this, and have come to the determination that it's largely confirmation bias and "prompt noise", similar to adding extra words in StableDiffusion to add additional (perhaps themed) randomness to the result, but isn't a means of exerting "control".

[foggy wing]

I see your point. When adding "scale" to the tags instead of just writing the name of the scale, I noticed a whole lot more scaled based arpeggios instead of just keeping to the notes of said scale.

[charred merlin]

Yeah, highly arpeggiated songs might have "arpeggio" in their "trainig tag data", but it doesn't mean that you're able to "tap into that", it just suggests that style instead of something entirely unreleated.

[unique stone]

It's possible that it's seeing the two different tempo numbers and doing something with them but not really understanding other than "they look different". It'd be really nice if it did understand and try to match the inputted tempo though.

🤔 wonder how hard it'd be for Suno to generate a metronome itself and then feed that in to the model as context, and then exclude that from the output

[charred merlin]

Straight to #1252298754787377225 with that one, dude.

meaning -- add it ❤️

[naive quail]

or just let us trim off the uploaded clicks

[charred merlin]

I will forward EVERY bpm related question there.

[naive quail]

...and the yodeling

[foggy wing]

https://suno.com/song/a51a11de-8222-4f01-94b9-ad031183f482
https://suno.com/song/742b3695-4a59-4f88-9dc6-7cb32958b914

The ONLY difference between these two is that I removed the keywords and just entered the values in the brackets.

[faint crypt]

With the upload audio option if I upload original beats and write lyrics in them can I make them public? Because this popped up and I guess its because it found vocals or something

The unable to make public thing

[unique stone]

I think that'll be possible eventually. I mean people want to cut the end off songs, being able to cut off the original upload would be a logical extension of that

[unique stone]

Maybe later 😆 /goes back to work

[hollow sapphire]

"I just did some trials to maintain the spirit of the musical style when I translate or adapt my lyrics for an English version. Until now, I was a bit frustrated because, even though the result was satisfactory, the style varied depending on... So, in this case, I'm putting the 3 versions here:" #1154779533627621518 message

[alpine venture]

what's up pinkimpala

[faint crypt]

So I have to upload just beats, with no vocals?

Yo

[alpine venture]

no, you can upload vocals

it just won't be allowed on public

that being said, you can still share it

and it can go on playlists

so like, if its good, and people see it

[faint crypt]

I know but I need it public doe

[alpine venture]

it can still get on commonly played playlists

[faint crypt]

But that can happen only with beats right

[alpine venture]

this only with beats, or non-vocals rather

[faint crypt]

So it only works with beats that's okay with me

Thanks

[alpine venture]

no prob; to be clear, it doesn't have to be a beat, it could be a guitar riff or something of that nature

but typically yea, people are using it with beats

Only time I've used it so far is with a siren

But I'll get around to it eventually

[faint crypt]

Yuh I get it thanks I just didn't wanna waste credits u know so I better just ask

This changes everything for me the songs gonna be a thousand times better

Wait if I only can upload 60 seconds and the song is needed to be 3 minutes I can expand it and then it will fill in the missing 120 seconds automatically, or am I stuck with a 60 second song

Dumb question I know but still

[inner socket]

Oh, my God😭 ! Indeed, I'm in a similar situation! I've been creating with v3.5 models until yesterday, and they can be over 3 minutes long, but yesterday the results were very short!🤕

[charred merlin]

@alpine venture, I think I saw a post elsewhere -- are you Canadian as well? 🇨🇦

[hollow sapphire]

I wonder if our own creations in Suno can be trimmed to a duration of 60 seconds and uploaded to retrieve a sound that works well? 🇨🇵

[charred merlin]

Are you asking if you can trim your Suno songs to add them back as Uploaded Audio -- you absolutely can!

[hollow sapphire]

Ok!!! Thank you so much, i'll try that!!! 👌

[charred merlin]

#🗳┃mini-polls message There's a post in mini-polls asking what features you'd like -- again, not sure why they didn't use a multi-voting poll, but if you react with A, it'll let them know you want to edit parts of a song directly within Suno so they know that's a desired feature.

But beyond that, you can trim your Suno clips in an external DAW and reupload them 🙂 I've had good results using the free stem splitter and DAW from BandLab to remove vocals from a track that I wanted as an instrumental, or desired different vocals on top of

[Here's a link to REALLY helpful BandLab tools so you can do all of this on the web, from your browser, without needing to install anything <3](#📚┃suno-school message) ... it's almost... too easy?

[hollow sapphire]

#📚┃suno-school message

[past aurora]

Hey everyone! Does anyone have any suggestions for tags that can be used for a duet with male and female vocals? Thanks a bunch!

[charred merlin]

I apologize that I'm not able to be of any immediate assistance about this -- but after running into a bug this reproducible, a lot of my "understanding" of the platform comes into question, and I'm not comfortable wearing a crumpled newspaper hat that says "expert" -- but I'm very happy to don a Deerstalker and Pipe, and start to reexamine the matter from elementary details 🔎

[charred merlin]

He's going to kill me for continuing to suggest he knows what he's doing, but @naive quail was doing a lot of duet songs for a bit, and he managed to get a lot of success. iirc, he says that it's entirely luck, but if anyone gets lucky this often, it's worth looking into what he does and asking targetted questions for him to follow-up on. https://suno.com/song/91dff5b1-6415-4b1a-8b88-25fd4ce1130a

This community is a wealth of knowledge, you just need to find collaborative partners to share ideas and results with.

Shout out to my Sunovator friends ❤️

[past aurora]

Many thanks Good man. I will follow up with him. 🙂

[alpine venture]

hey, thanks for the headsup, I'll look at this when I have more creds

its always nice to find new methods

[alpine venture]

Oh, no, I just have some canadian friends in here

And I mean, a lot; most of my spotify plays come from canada

well, not a lot haha; aht I mean is, for me its a lot

[charred merlin]

Heck, even in just your approach to SD you're outrageously creatively innovatively swell-- I'd kill to get your input on my recent post, where I outline a method for "turning suno on a dime"? https://discord.com/channels/1069381916492562582/1256301881966198805

"intro", "terminators", "pivots" -- perhaps an attempt at throwing spaghetti at the wall, but I'd say that intermediate > advanced users are about "cooked enough" that being able to use our own "jargon" could be helpful to clearly communicate ideas and build off each other?

9am AST, cat on my lap, coffee in hand, Suno trending playing in the background -- NOW it's time to start 😉

[charred merlin]

Again, it's one thing to point to a specific coupling of songs -- and while having directly comparable styles prompts and lyrics used between makes it a LOT more meaningful, without testing how Suno responds to specific BPM requests (which we have, extensively, to no avail), it's more of an interesting data point to contribute to the on-going conversation, however I don't know if it highlights anything that we've not seen before.

It's also not fair for me to respond so dismissively, as my intention isn't to dissuade you from experimenting and "getting a feel" for how everything works -- and it would be absolutely unreasonable for us to ask you to try to follow along with months of interspersed snippets of dialogue from the past few months, but I WISH we could "code" or "command" Suno reliably in that way, but again... everything I've seen only further dissuades me from "wasting credits" on a sunk-cost fallacy.

Genuinely appreciate that you're following up with examples to compare from -- I absolutely notice the difference, though if I distilled a takeaway from it, it would be that some prompts are more strongly weighted for getting specific styles, and it sounds like one of them "understands the assignment", from a "blastbeat" perspective 😛

[foggy wing]

I agree and I do not find you dismissive, I also understand us new users are reinventing the wheel. But also wasting credits in doing so.
Sometimes I just would wish for a more structured advanced tab where we could set different values and the generation would take place upon the template we create.

[faint crypt]

Wait if I only can upload 60 seconds and the song is needed to be 3 minutes I can expand it and then it will fill in the missing 120 seconds automatically, or am I stuck with a 60 second song

[charred merlin]

In regards to wasting credits, what I advise is to do as much experimentation as you can near the end of your daily/monthly credit cycle before the credits roll over.

Beyond that, I've been assembling "Prompt Experiment Playlists" over in #1256301881966198805 (though admittedly I've had to refocus efforts elsewhere, so I've been delayed in publishing my findings as quickly as I was hoping to ❤️ )

[charred merlin]

You Extend the song afterward, and use "get whole song" to stitch the results together -- by default "Extend" adds onto the very end of the song, but you can choose WHERE to start "overwriting" the track instead of the end.

Since it automatically starts at the end, make sure that it's not extending from silence, or the extensions will likely only be a couple seconds or two.

[trail acorn]

Suno almost nailed my vision in one try 😂 But it didn't listened to my suggestions and switched the tempo and expression in verses and choruses. And i don't know how to tell it to add rattling sounds... Do you have any tips for new user?

https://suno.com/song/11a5e676-17e1-4363-bab2-c828fb22b035

[charred merlin]

Hey @trail acorn! Funny enough, we were JUST talking about how "tempo" is interpreted as a "thematic suggestion" instead of a "command / code" -- let me link the conversation ❤️

#📚┃suno-school message

[alpine venture]

Ooh, I'll check this out

[charred merlin]

There's lots of helpful tips in this thread; it's a great place to read up on existing conversations, (just do a search for "in: 🔧┃technical-discussion" ( it should autocomplete, or just hit ctrl + f in here if you're on desktop, then search for whatever terms you'd like to read up on!), you can find collaborative musical partners, and ask questions about how others approach their use of Suno so you can better inform yourself.

Or you're welcome to come keep me company in #1256301881966198805 and give me ideas on what else to test -- LET ME SPEND MY CREDITS SO YOU DON'T NEED TO WASTE THEM before you get a feel of how a prompt will work before using it in your song!

[alpine venture]

I'm out of creds atm so I can't be much help in testing and seeing, but I'm excited to check it out

[charred merlin]

oh my dude, even giving me "snippets" of what to test helps.

https://tenor.com/view/the-office-spill-chili-pot-kevin-malone-gif-3915565

me with credits -- I need your help using them before this happens.

[trail acorn]

Is there a way to, uh... push it more in the direction i want? I've used https://discord.com/channels/1069381916492562582/1261146873331318835 tips and tricks and i'm not sure if i wasn't as descriptive as i should be 😬 please @ me if someone knows : )

[charred merlin]

#1256301881966198805 message

"Terminator Extensions 😎
aka. how to make your song stop doing the thing, and start doing that other thing instead"

[trail acorn]

Oh, thanks. I'll soon look at it 😄

[charred merlin]

If you end up making any sense of my ramblings, and are free to even spend 5 minutes chatting in DM later about your experience trying to nudge the "rolling boulder in specific directions" before and after trying the method of "terminator extensions"?

[faint crypt]

It repeated the verse 3 times

Nvm I'll figure it out

[charred merlin]

#📚┃suno-school message It's not uncommon for Suno to repeat verses as it tries to fill out the entire song length -- instead of trying to force it to listen with hyperspecific prompting or metatags, I find better results placing extension points directly before I want to change the song -- that way the final result is a continuous flow of all of the "moments I wanted to keep"

[trail acorn]

English is not my main language, so i'll need some time with deciphering it 😂

But i'll come to you if needed 🙂

[charred merlin]

This is very helpful as well! 🙏

I don't want my guides to be "tricky" to understand -- and I may speak "too colourfully" at times -- I don't want to make things harder

There is no rush or obligation to let me know, but if there are sections that are too confusing, or if you believe it would be helpful for it to remain "wordy", but you believe I should look into translating the guide, it would be very helpful ❤️

Thank you for your time, @trail acorn, please say hello to @uncle when you see him next ❤️

Toker, I don't know if I got your opinion on the terminator-extension thing? I know you're an old-hat pro, so it's not about learning a new method per se, but I'd love to know how someone like yourself sees it in terms of whether it's helpful from a mentoring standpoint to new or casual users?

[naive quail]

grumbles

Those Fallout duets were really just happy accidents, tho. I'm convinced that Suno turned them into duets because of the lyrical content. First person perspective, the use of "We, We're, We've, Us", etc. After the first one, I tried to take advantage of that within the lyrics with "We're the Atomic Duo" and stuff like that. 'Seemed to be effective?

The only one in my Duets playlist ( https://suno.com/playlist/12529891-cfa7-4302-8c68-f4f219cd96b4 ) that I specifically prompted to be a duet is Echoes in the Abyss and that one took a number of extensions to get the parts right. I have gone back and edited the Fallout ones to better reflect what Suno actually did with them.

[charred merlin]

SUCH an important point that I always forget to highlight, is JUST how much the lyrics can impact the final result.
https://suno.com/song/65a2e520-e54a-4def-be67-d483a08a49eb -- no style prompts entered, just lyrics

[faint crypt]

How do I place the extension points, and how do I make so it says what is written only and how do I make it so it's only 2:30 the song is instead of 4 or 3:45.

[charred merlin]

Hey @faint crypt 🙂

The elegant @final hazel made this wonderful 4m guide for getting familiar with the basic layout and tools of Suno.com, it shows how to get started with Extensions, piecing the resulting "part 2, part 3, etc." generations into a single song.

[faint crypt]

Thank you

[charred merlin]

(sorry for the double ping daizy, 1 more and who knows what floodgates might have accidentally been opened 🤣)

[alpine venture]

this is a great idea, thanks for taking the initiative to do this

[charred merlin]

You're very welcome!

In general, I find that the more you ask Suno to handle at once, the more likely it is to "get confused", "ignore" or "downgrade" how heavily some prompts are weighed compared to others being used at the same time, etc.

While some of my best results have been "one-shot generations" that I made without a lot of pressure, the best advice I can give when you're trying to achieve a specific outcome is not to try to handle "everything all at once" -- try to effectively use extensions, extending-from-extensions vs extending from "get whole song" -- just experiment and "Sunovate" 🙂 If you feel like Suno did a goob job up until "1:37", then go back a couple seconds to "1:35" take out all of the other style prompts and lyrics (it will continue from where the song left off, using that as context, so it probably already understands what styles to use, but you can "remind it" if the extensions get a bit wonky), and try to focus on finishing the rest of the song beyond that point (this means that no matter what happens, the first 1:37 is "locked", and you don't need to worry about wasting 1000s of credits trying to get SUPER lucky and get everything done all in one try -- you'd be better off spending your time in Vegas if that's your usual way of success 🤣 ❤️ )

[faint crypt]

Thanks bro I really appreciate it fr

[brave urchin]

probably been asked, but are there plans to offer options of uploading a voice profile to use? For instance... I can carry a tune a biut... would be cool to have "me" sing back the songs rather than making instrumentals to sing with.. .... Wife's request. not mine

[final hazel]

The only way atm is with an audio upload and extend from that. Here an example of mine: https://youtu.be/6rzt1WNGUSc?si=vg9DlCLgfUtR6Aom

(it's was a sneak preview before it became available)

[brave urchin]

I'll check that out. thank you.

[charred merlin]

Hey @brave urchin!

https://discord.com/channels/1069381916492562582/1252789683176345641 we have a known-suggestion for adding reusable vocalists, so people can make an album sound like it came from "the same band". If you upvote the linked post it will help the developers know just how many people are itching for that feature.

If I understand correctly, beyond wanting to reuse vocalists that you generate in Suno, it sounds like you'd be interested in some form of vocal-model training, similar to how ElevenLabs and their ilk handle voice cloning?

I hope they add that soon, but if not, I'd like to tell "that idea" directly...

https://tenor.com/view/jim-carey-good-afternoon-good-evening-goodnight-i-dont-see-you-gif-15609107

[brave urchin]

I'm not familiar with that service, but yes, a "sampling" of vocals for use as model of the voice in the music or even just a chorus or backup vocals. just depends

[charred merlin]

I'm not SUPER familiar with it either, but my iPhone had a voiceover feature I read about that was introduced in one of the newer iOS versions -- you spend like...30 minutes speaking into it (they let you break it up into multi sessions, thank lord), and afterward it "learns" what your speech pattern is so your phone's text-to-speech can use your voice.

I'm sure there's legitimate, appropriate, mature uses for it -- I just use it to "shout" random unexpected words while my wife is focusing on something, then attempt to claim innocence because "I didn't say it, technically".

[sturdy cobalt]

asong agey

[safe hazel]

What is this assertion based on?

[charred merlin]

Hi @sturdy cobalt!
(my best friend is named Troy, so I'm extra happy to help you ❤️)

If you want to make a song, Discord doesn't work any more 😦

This site is where we make music now :)

This video can help you learn how! (more languages and other resources available here https://discord.com/channels/1069381916492562582/1151542418991239168)

[sturdy cobalt]

genarate

[charred merlin]

NOT ON DISCORD 😦
🤖 ⛔ 🔊 🎵

MAKE MUSIC HERE

[safe hazel]

If you have a string of extensions for a song (like Part 2, and Part 3 etc) and you only want Part 2 and Part 3 (but not part 1) is there a way to do that? "Get whole song" will include all parts.

[naive quail]

Not at the moment, but it does get brought up a lot. Hopefully we'll be able to do it at some point in the future.

[safe hazel]

Thanks for letting me know!

[slender stream]

i have a question. the wav files you download from Suno, are they lossy or lossless?
by lossless, i mean the frequency on the spectrum goes all the way to 22 kHz with no cut off.
thanks

[trail acorn]

Does anyone know what metatag i need to use to make some part of song last for example: 10 seconds?

[foggy wing]

Try [short interlude] or something like that. I've seen people use double dash "--" as terminators also, but haven't gotten it to work really myself.

[trail acorn]

Thanks

[charred merlin]

https://www.suno.wiki/faq/polish-and-production/can-i-download-uncompressed-audio/ I'm not entirely sure, but it looks like they've stated that any perceived degradation in audio quality is "baked-in" as a part of that particular generation, not as part of compression -- so I feel like the answer is lossless?

[valid dagger]

I in fact tested it. Downloading the mp3 will give you lower quality, in particular if the sound quality is not the best to start with (like these out-of-phase drones V3 does on multiple extends). Downloading wav gives you 48/16 with all the quality the AI originally produces. Quality has noticeably improved with V3.5

[daring nebula]

Will there be an option in the future you can also download the songs with all different stems? It would make it easy to edit that to get the song top notch with mix / mastering.

[hollow hare]

You can vote for such a feature at #🗳┃mini-polls message - or at least separating the vocals from the instruments.

They devs have tried it in the past for v3.5 but the output was worse than running through a stem splitter by other parties.

Sounds like v4 (or a later version) may try changing the approach to try to make a better stems output one day? Time will tell

[daring nebula]

alrighty, thanks for the answer

[hollow hare]

No problem. I should point out that the main issues with providing stems is that the training the AI went through (likely) learned on a mastered audio file, not individual components, thus the AI generates music like a mastered track based on the sum of the waveforms (well, not quite, but gives you an idea)

Hence why the poll is about simply separating the two-part vocal and instrumental instead of vocal and individual instrument components (which likely such involved splitting will not be available on the platform).

Though, mind you, I am not a staff on the team. Just a person who lurked here and helping people time to time

[daring nebula]

I still appreciate the answer, it already gives some intel that at least they have tried it out / have the option in mind.

[viral lotus]

@charred merlin I have the perfect generation, I just can't get it to end when the song ends. It should end at 2:59 or 3:00. I've tried v.3 with "[End]" in lyrics and "End" in style. Any other suggestions? It just starts right back up for a one minute+ encore. https://suno.com/song/944e70c0-7a98-4ec6-b0ae-5c9a06d4a2c8

[naive quail]

Have you tried extending from the extension that starts @ ~ 02:40 before combining the sections? Sometimes that'll work better than trying from the combined extensions.

You could used Get Whole Song on this one to pull it back into your library. It's 00:06 in length and I think I did it from 02:57 or :58. There is an extra "Your infirmary" in there, tho. https://suno.com/song/0e70e6e4-47a0-455f-b789-ee9448f0ea50

[brave rain]

So, last time I generated music, I was getting short songs. This time I get a full 4min generation but 45 seconds of silence at the end. sigh

[brave rain]

Oh, now back to short generations. 1:30 using v3.5 is just bad.

[rare oak]

At least it's not some weird track parts popping out of nowhere

[brave rain]

Oh, I got the short, split track too.

Less than two minutes, two different "songs"

[fickle storm]

Is there any way to stop Suno from soloing over everything when using instrumental? I feel like I have tried everything under the sun to just get a rhythm section

[rare oak]

You tried an Interlude tag ?

[fickle storm]

Yes. It seems to just completely ignore tags

[wet minnow]

try to give it a load of metatags to work with ... intro, bridges, buildups, drops... everything you can come up with to keep suno from using vocals

[fickle storm]

Is that how it's treating Instrumental? Just replaces vocals with leads/solos?

[wet minnow]

I think that is whats happening. I imagine the voice as a synthesizer, and suno decides if the melody is played by voice or by an actual instrument. But thats just my feeling.

[fickle storm]

It definitely feels like it does

I'll mess around with tags and see what combination distracts it enough

[unique stone]

I have a hallucinated vocal part that Suno generated which was very noisy so I played the melody of it and mixed it together, reuploaded it and Suno and typed in the phonetic sounds of the "lyrics" - Suno is able to follow the original melody with "lyrics" as a female vocal, male vocal, and also has on several occasions just played an instrumental version of the melody. So it seems able to recognise what part should be the vocal and treat it either like a vocal or instrumental. Pretty clever stuff. Wish I understood how that worked 😄

[swift zealot]

I have been using hallucinated vocals to make most of my songs, but now it seems that it wont generate any hallucinated vocals any more. Does anyone have the same experience? I am using the same keywords as before. Did something change or get updated in the AI?

[untold briar]

How were you generating them before? Generating hallucinate vocals can be done in many different ways. The simplest possible way is to use no lyrics, but add "vocals, vocals" to your style prompt.

[sleek gazelle]

Anyone getting this error too when trying to download a WAV file? Worked fine a couple of days ago.

[swift zealot]

Yes this is what I did, simply adding the vocals tag to the prompt

[untold briar]

Hmn, I wonder if Suno is getting better about not singing vocals in instrumentals. If that is the case, you could use the other methods (unsingable text like morse code, [incomprehensible], random made up words, unparseable lyrics).

Let me check quick. You could always stack on additional vocal tags to the style prompt, brute force too.

[neat nexus]

Seemss to be running fine for me.

[swift zealot]

Thank you for checking! I have been trying to brute force it but nothing seems to be working. I had one tag that worked, but screamo does not sound that great over every song

[untold briar]

I didn't check that just one or two "vocals" wasn't working, but this works: "indie folk, vocals, vocals, female-vocals, female"
https://suno.com/song/66133910-f267-4154-b46d-a7b13f054b5d

[swift zealot]

hmmm these tags seem to be working for me as well. Perhaps its just the hardbase rock genre

[untold briar]

There might be less diversity in the hallucinated syllables, more repetition. If so, this is sort of 'better' in that Suno is basically sticking close to the 'actual' lyrics. Even if people like you and I would prefer otherwise. Ohh, but I still got some good ones. I think it's mostly still good.

[untold briar]

I did add both "vocals" and a gender "female", since female or male almost always refer to vocals adding them can also help

[untold briar]

Another thing that can help is to add just a tiny bit of lyrics. Can be one or two real words, or just random characters.

[sleek gazelle]

Hm, weird.

I'll try again tomorrow, thanks

[swift zealot]

Thanks for the tips! I will do some experimentation and see if I can get it to work!

[swift zealot]

With the double vocals tags hardbase rock, vocals, vocals, female-vocals, female i got it to work again! Thank you so much for the advice! I think that something must have been updated since the previous tags did not work any more.

[untold briar]

I do feel like Suno is trying harder to not sing words you don't have in your lyrics, possibly. So even when you have lyrics like this, rather than "SimLish" (words that sound like a fake language) Suno mostly tends to sing oooohhh and ahhhh and humming instead. Not always but more often than I remember.

https://suno.com/song/bef67709-cb0a-4428-a40d-bac61a878cba

[untold briar]

You can also use a regular verse of lyrics, then Extend that verse with no lyrics (or morse code, or whatever) and Suno will either repeat verses from the previous clip, but also can hallucinate variations of verses that are similar but not the same as the previous lyrics, which can be neat.

[swift zealot]

Interesting observation!

[swift zealot]

I find Suno to be a bit less free when I do this, but I have not tried to add morse code or noise to the lyrics box yet, something to try!

[untold briar]

I can't speak for 3.5 right now, but in the past it was pretty free. One of my fav examples: https://suno.com/song/f193f729-cd82-48e9-8454-39e48235e3cf

[swift zealot]

ah yes this is great

[heady sun]

I have a question:
How do you deal with songs that generate a pause and then pick up with a musical outro?

Song in question at about 2:20: https://suno.com/song/ff85588d-7949-49b4-acab-93ad76a62871

P.S. I don't mind songs that continue with a musical outro.

[mystic flume]

Define "deal with" -- if you don't mind it, how are you intending to deal with it?

If there's a pause and an outro I don't like, I edit it out in Audition.

[mystic flume]

I've gotten some really good results by just letting Suno sing gibberish, and then transcribing that gibberish to get the expected rhythm of the lyrics down and writing my own lyrics to fit that meter.

[untold briar]

Good idea. When you Extend the gibberish with the lyrics that fit the meter, Suno matches up the previous rhythm as you expect?

[mystic flume]

Well, the answer to that is always "sometimes... eventually." Let me see if I can find a good example.

Hmm... I don't organize stuff well enough for that unfortunately. Like, even the gibberish I just write over it instead of saving it.

[untold briar]

I wonder if you could help Suno by 'anchoring' with a few real words in a larger prompt full of unsingable gibberish, and it would be able to align things up better when converting into full real words.

[mystic flume]

Frequently my process involves bending a generated instrumental to my will for hours before finally producing some final lyrics... and then I end up with a much better result doing a full new generation with the same prompt a few times.

Figuring out what sounds good, and how Suno interprets things, is a big part of the process for each song.

[untold briar]

Haha, I feel that. Just in general it's way too easy for me to get stuck trying to Extend to fix or finish a song, and then I just start over and get something better almost immediately.

[mystic flume]

Yep, because you've charted a path with your lyrics and structure that Suno can easily follow.

[untold briar]

Yeah, there is value in going with the natural flow of Suno. You don't fight against it.

[mystic flume]

That's one reason I wish I could get the Seeds, because it would be really cool to be able to do that, and then just do an entire new generation with the lyrics using the same seed to see what happens.

Though I understand what a disaster revealing the seeds would be.

[untold briar]

I have felt this way in Bark, so I'm not sure I'm seeing it in Suno unbiased though. In Suno's Bark TTS, if you try to make music, it's incredibly hard to also prompt the exact lyrics you want it to sing. But much, much easiser if you allow Bark to hallucinate while singing. So I came in Suno Music with that feeling.

[mystic flume]

Even just a "use the same seed" option would be nice.

[untold briar]

I think seeds might have less than other AI if I had to guess. At least in Suno Bark, if you use a seed but change the prompt even just a few words, it's almost like you didn't use a seed at all. I still felt that it would still be worth having a Suno seed, just because people want it and it would still have some edge case uses though. But I can understand why it's not a priority. (I think it might also make things complicated on the server side, but I don't know.)

[mystic flume]

ok I don't wanna brag but I just made a script that will split a Suno wav into R and L tracks, cut them into 30sec clips, run each clip through Audio Super-Resolution and then recombine them.

[untold briar]

I only tried it a long time ago when it first came out, but I found using overlapping audio chunks had weird artififacts. And not overlapping audio had other artifacts. But I can see it's been updated since then, quite a lot...

[mystic flume]

Can I tell the difference with my ears? No. But the spectrograph speaks for itself. (I still have some kinks to work out because there's an issue at the very end, but so far so good.)
Before:

after:

[untold briar]

You can crank up the guidance and steps a ton to make it more obvious

Maybe that's why I had trouble with long audio in chunks. As usual, I crank things up to 11 right away lol

Ohh, one thing I tried was downsampling the audio first, before upsamplling. In some cases it was better

[mystic flume]

I did gs 7 and steps 100 for this. The volume drops off at the end in that spot where the graph suddenly changes. Might be due to the fade out.

Once I finish it I'll put the script on github for anyone who wants to fiddle.

[viral lotus]

Thank you, I'll give it a try. I've been starting generations nearly at the end of the song.

[pliant pumice]

hunnytheurry yes but often the song are short then 2 minuts even if I respect all the rules 😭

[merry creek]

Mobil does not seem to have a way to make a song a full song. Is that capability for paid users?

[mystic flume]

Mobile browser?

Subscriber view:

Free user view:

The option doesn't show up until the extend finishes generating, though, so maybe you just need to wait.

[merry creek]

Okay. Thank you.

[hollow hare]

Cloudflare explains the 524 issue in the bottom right of that message. Just refresh the page in a few minutes:)

[mystic flume]

Do you find increasing the guidance actually improves the audio quality?

[untold briar]

Higher guidance values messed up music, but were good on drum samples.

From memory, the thing that had the biggest effect was upscaling twice. Upscale once and it adds a faint bit of audio between 16 and 20 khz, and then do it again and that faint audio got fully fleshed out.

I don't think downsampling first was good, but I was testing mostly samples, not songs. Not sure I even tried a voice with that. With something like suno audio vocal degradation, it woudl sort of vaguely make sense that maybe downsampling could be good though? Give the upscaler more room to make changes?

[mystic flume]

Interesting. Yeah, I just tried gs 20 and it sounded terrible. I'm also going to work on upscaling once, then creating stems, then upscaling the stems and recombining.

No idea if that'll do anything. Gonna try your suggesiton of upscaling twice first.

[untold briar]

Are you using overlapping audio when you split it, or are the chunks fully discrete?

[amber vigil]

if i wanted to mimic a different style (basically sampling) in a suno song, i know i can extend and change the style tsgs and get a slightly different sound that still synchs with the original part.

(Assuming i have that right)

But can i then extend from that new part (part 2), switch back to the original style and wrap up the song so essentially, i end up with a song that sounds the same except for the “part 2”?

I’m tired atm so apologies if that question needs to be restated for clarity.

[harsh island]

On a sidenote, not meant for an entire discussion, and probably asked 10^256 times.
Any reliable way to add in a sound effect liek record scratches, single church bell tolls, etc?

[untold briar]

Record scratches are pretty reliable. Lots of different prompts work like [vinyl scratching] and [turntable-scratches] though I usually also have terms like this in the style prompt. Not sure about church bells.

[harsh island]

Yeah I kind of want to make something that's like

[Church bell toll] lyric line 1, 
[Church bell toll] lyric line 2.
[Church bell toll] lyric line 3.
[Church bell toll] lyric line 4

But only in that section of the song.
Can I I download it, add them from external audio software, reupload, and extend from that?

[untold briar]

Yeah that will work pretty well. But you only have a minute of audio max in Upload.

If you Extend the song, and try to prompt [bell] again, it may or may not work and summon the sound you added. It can work for sure, but that isn't reliable.

[charred merlin]

I’ve noticed that sound effects are most reproducible if they’re reasonably found in music — sirens, lightning, censor bleeps, I’ve even heard the occasional gunshot in a song or two

[harsh island]

I found the censor bleeps yeah.

Hurt my ears

[untold briar]

Oh is there a way to prompt for a censor beeeeeep where it sounds like the song was sensored? Like maybe you hear just the F sound then beep? I can't believe I never tried that. Sounds hilarious.

Like you can hear a little bit of the word that was censored behind the beep?

[harsh island]

Not sure it's that fluid

[charred merlin]

Yup — I think it was something Toker or Hamster threw together for a joke, so I don’t recall the specific way they used it, but I’ll try to find it to play around with the idea if you think people could have fun with it

[harsh island]

This is fine, its for an intro, outro, bridge or something else that's just a once-occurence

Any software that's cough free cough you can recommend?

[charred merlin]

Actually, I just remembered how funny Kimmel’s “This Week in Unnecessary Censorship” was — this would be comedy gold.

[harsh island]

Ahaha, omg yes I would love to hear somebody make that

[charred merlin]

I’m about to pass out and my brain is running on fumes, but what sort of free software are you looking for? Not a sketchy question to ask if there’s a free alternative to something ❤️ 💸

Money is the absolute worst—all my friends hate money.

[harsh island]

Something... that doesn't require a sound engineer degree I suppose

[charred merlin]

Jfc — @untold briar are you an alpha tester?

[harsh island]

Don't knock fumes, they can make your brain go brrr

[charred merlin]

Ugh, no. Doesn’t look like it?

[untold briar]

I lost the thread. For Audacity?

[charred merlin]

If there’s any alpha testers on, ask them to go into my old “common repost thread”, copy the one labelled “BandLab <3” and share it with Eardrum

Actually, if someone could DM all of those links? Not many, but they’re helpful to me

[untold briar]

I found it in search: #📚┃suno-school message

[harsh island]

Thanks 😄 Gonna check those out

[charred merlin]

@harsh island ^ this website BandLab is about to change your friggen life bro.

DAW, AI splitter, AI-gen samples to start your song, royalty free samples, collaborative social media platform, licensing on the way

AI auto-mastering

[harsh island]

I actually started looking into more actual song engineering -- since Suno doesn't always 'exactly' does what I want it to do.

[charred merlin]

Legitimately all free forever, unlimited usage on free plan.

Membership is only for “unnecessary extras” that you may want to grow into, but I haven’t needed to

[harsh island]

But it's a lot better than having your music based on 'ultimate guitar pro's real sound bank'

[harsh island]

I'll deffo check that out, thanks

[graceful condor]

What the beep

[charred merlin]

Oh, I must be running on less fumes than I thought — I thought you were looking for an external free piece of software.

My bad ❤️

[harsh island]

I was

But I'm not knocking useful tools and software ;)

[charred merlin]

Can you elaborate on what you mean by song engineering?

You can create beats and loops in BandLab , or you can import songs from Suno, separate the stems (say to remove the vocals) then trim out a cool sounding section, then upload that to Suno to use as an audio upload?

Unless I’m misinterpreting the question, BandLab offers a full free digital audio workspace — but there’s paid alternatives, like FruityLoops, etc.

[harsh island]

Well I have a friend who does semi-amateur music production, and he just tells me to get these mix-tables, then throws up things on the screen that look like you need an engineering degree. That's why I mentioned that term 😅

I think BandLab will suit me just fine for now hearing that, I can't check it out right now due to work.

[untold briar]

If you take your song and just stick church bells in front of the start of it (not even on top). Then Extend. Suno can even work with that, sometimes.

[harsh island]

Nah I think I want something more specific/micromanaged than what suno will be able to give me with prompts

So thanks 👍 i'll deffo check it out

If I get it right, will post the result in the showcase

[charred merlin]

Even if it doesn’t turn out perfect, we’re always down for hearing experiments 😄

Even if something doesn’t work, you can still learn from that. And maybe it could inspire someone else to try something similar!

[untold briar]

I use Davinci Resolve for audio editing, since I have it installed anyway for videos. The free version is very powerful, though it is annoying laggy with audio edits.

[charred merlin]

Kinda like how I felt like Super Amateur whenever I’d see someone talking about their spectrographic analyzers, it felt super out of my reach.

Luckily a collaborative friend was brave enough to share their findings after doing their own research into the matter, and now we both get to be a little bit less clueless about it each day.

There’s almost half a million people on this server, if you put yourself out there “on a creative limb”, you might find a collaborative partner who pushes you to try new things and be better 🙂

[harsh island]

It would be really cool to actually get my own guitar-riffs into suno ngl

[fluid robin]

for the ones that like to do rap songs put this in style of music Static Rap. Dubstep. Vocal Dub beat

[fleet meteor]

Probably has been asked before. Do you get to keep your credits at the end of the month if you don't renew the subscription and haven't spent more than half?

[brave gull]

it kinda works but depending on how you write it sometimes suno drops the n-bomb even though you didnt write an n lmao

[gleaming briar]

How can i contact support? I need to cancel a subscription on a lost account.....

[untold briar]

Email [e-mail]

[untold briar]

Monthly credits always go away at the end of the month. Or if you are still subscribed, they are reset back to 2500 or 10,000. If you have 'Extra Credits' (it will show on the subscription page) those credits do not go away. You might have some because of giveaways in the past.

[sweet wigeon]

Is it possible to use this to change instruments?[Switch from violin to cello.]
[Switch from flute to contrabass.]

[brave rain]

[cello solo] maybe

[sweet wigeon]

begin of the song was happy A Major for violin. Now I need to switch to dark scales C Phrygian. So I'm telling Suno, to switch from flute which stays in the upper tones. I cannot have cello (bass) and contrabass solo, there must be other things like piano darker or drama tones.

[brave rain]

I don't know if Suno allows sophisticated enough control to achieve what you want, but good luck.

[sweet wigeon]

Yesterday I god satisfactory results the mood changed from happy to dramatic, dark, death like, then melancholic. But the voices where whispering. Today trying to do the same, but with 50 credits per day cannot do much.

[charred merlin]

There's been some really creative uses of the audio upload feature, as that's currently the only way to 100% guarantee how your song starts, but I understand that it doesn't "scratch the itch" completely.

As it stands, if you can include a guitar solo within the first 6-60s of your audio upload, Suno might be able to "hook onto" the music style and continue it, or based on the prompt suggestions you choose, it could provide a reasonable rhythm guitar line to complement it. I wish I had more actionable advice, or examples to point to, but audio upload is so overpowered, that it almost feels like cheating 🤣

While I don't think it's even seen the light of day, a lot of people talk about potential features in the next model -- "inpainting" usually gets brought up as a means of letting Suno "correct" or "reimagine" specific sections of a song, but you just gave me a phenomenal idea of combining audio uploads WITH inpainting -- allowing you to record your own vocals or guitar solos and ask Suno to cohesively mix them together

[harsh island]

I'm just waiting for a feature that allows me to extend from an extend without having to include all that went prior to it when collecting 'whole song'. Sometimes it starts a new song at the end and it's straight up fire.

Maybe cut them too 😂

[charred merlin]

You CAN extend from an extend! That's a way of ensuring that you don't use the earlier section of the song as context -- extending from "get whole song" ensures that any further extensions are able to "see" the earlier context so it can attempt to continue and musical passages or motifs.

[harsh island]

I did that, and then when I went to part 3 and got the whole song it gave me 1 to 3

[charred merlin]

Yeah, builds upon each extension that you add (depending on what timestamp you started it from)

#📚┃suno-school message
Here's a guide I wrote on the concept of "terminator extensions" -- using a 1 second extension to effectively set the "context memory" to a single second for part 3 or beyond -- effectively allowing you to prompt the song in any direction that you want.

[vast tangle]

Test

Guys, I have creating problem in suno, can someone watch this video and help me why it happens?

[charred merlin]

The message you're receiving at 0:24 is only intended to display the very first time that you click on the "custom mode" toggle.

Without further context, it would seem like there's a browser issue -- does the issue persist if you try in a private window, or a new browser? (it's important to make sure that the site works by itself without any extensions at first, etc, so we can try to isolate what specifically is causing the issue)

[vast tangle]

I'll try it with chrome, that one has no extensions exists in it.

[vast tangle]

Ok so, I tried with chrome. I'm pressing create, but it doesn't works for some reason. No alerts or anything like that...

[quick geyser]

Will you improve the accents of different language or are you working on it?

[charred merlin]

That's definitely weird -- sorry I'm not able to be of further assistance, it might be a good idea to share a detailed report of what you're experiencing to [e-mail] ❤️

[vast tangle]

Thanks very much mate, I'll definitely try it 😊❤️

[neat nexus]

Almost half a million people here. Wow. That really puts things into a different perspective for me. sorry for kicking this up hours later but I had to comment and amplify this.

That's amazing!

[trail badger]

Security confirmation error in the member login section? Is there a general problem?

[charred merlin]

I know it's a bummer that they don't stockpile -- not that it changes the fact, but it seems as if AI-credit-based subscriptions often work the same way.

I used to compare it to paying for 1 month of Netflix, but only watching 1 week and feeling upset that they still charged for the full month, but while similar conclusions can be reached, it's ridiculously dismissive in hindsight -- especially considering how sensitive of a topic finances can be.

If it helps put anything into context, I now understand it more as a means of trying to balance server-loads to make sure that everyone is able to use the service fairly -- if the devs know how many free users they have, they can reasonably predict the "base server load" that all free users spending their daily 50 credits would add. Next, they think about all of the paid users, if they decided to spend the entire month's worth of credits all at once, they now need to think about that, but also make sure that the server stays light enough that paid users can always use the service.

It is very hard to predict how busy the server will be (though it can help to keep aware of when certain parts of the world have their peak usages), but by separating the credit distribution specifically month to month, or day to day for free users, they can provide better service for everyone.

(imagine if a bad actor or competitor saved up years worth of credits, then spammed the website constantly, almost DDoSing it 😱 )

The only actionable advice I'd have is to keep aware of when your credit cycle is drawing to a close, and commit yourself to experimenting as much as you can before they "roll over" -- for me, weeks 1-3 are for personal projects, and the last week of the month is when I go hogwild with experimentations and offering to test out prompts for other people.

That being said, maybe they'll address how credits can be stockpiled in the future if it seems feasible ❤️ #💡┃suggestions-web with such a great idea!

[charred merlin]

I'm not experiencing any issues signing in on the browser -- where are you receiving this error? Could you share a screenshot?

[trail badger]

"Cloudflare" security verification, "Verify you are a real person" fails.

[charred merlin]

lmao, Love you Woodhouse

[charred merlin]

[I've seen other users experience the same issue https://answers.microsoft.com/en-us/microsoftedge/forum/all/cloudflare-keep-looping-to-verify-as-human-on-edge/c94c1303-84eb-45d4-932c-6d4d1cfdc5d6), and it often seems to be related to adblockers, extensions, or the browser requiring an update.

[charred merlin]

Rogue, search for my “terminator extension” comment above ❤️

#📚┃suno-school message

“The pivot” would simply be “end” in the style prompt, no lyrics

[obsidian tendon]

Question about generating lyrics, does it always default to 2-4 word verses and chrouses. How can I get it to genenrate complex long lyrics? Not the same pop rhyming BS

Does it take into account the Style of Music when generating lyrics?

[charred merlin]

Are you using Simple Mode?

You don’t have a lot of control over the song this way.

[obsidian tendon]

Nope. Doing custom

I guess this is where years of being a singer/lyricist surpass even AI tools?

[charred merlin]

Oh, you can choose whatever lyrics go into that section. If you find that it’s generating too short of stanzas, you may be able to specifically ask it to generate longer stanzas before clicking generate

[obsidian tendon]

Oh I see. I can interact with it like a LLM. It will under stand instructions. Huh I didn't think it worked that way.

[charred merlin]

100%, but also knowing specifically how to use the tools is a big deal — the fact that you’re presumably a lyricist makes it even THAT much more powerful.

For a fun followup to this meme 😛 #😜┃suno-memes message

can't you command it

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

"you don't code / command suno, you talk to it."

[obsidian tendon]

Got it. Thanks. I am not a lyricist. I think I have decent ideas but unsure how to create the lyrics for the song that are good. Not low effort rhyming and catchy stuff. That is good from time to time but listening to some of the complexities in the music generation. I'm like how awful is the vocal range creation. I understand the dataset Suno is using comes into play with that. However, I was curious if we could control more of the lyric generation than throwing in a few sentiments and hoping it comes up with something better than a rhyme

[charred merlin]

If you're past the point of frustration, DM me the song you're working on and I'll show you specifically how I would try to handle it, and give step by step instructions ❤️

If you're working this hard on a song, it has to be GOOD

(just give me a shout out in the liner notes of the album when it goes multi-plat 😉 )

[charred merlin]

I know my perspective comes from struggling with v3 not singing my carefully written lyrics ANYTHING close to what a sensible artist would interpret it as, but v3.5 is a literal mind-reading god at producing cohesive lyrical flow if you have tangible lyrics you're trying to generate -- 9/10 times that I have a reasonable idea of how I personally would sing the lyrics, it predicts the exact timing and cadence. (it does it's own thing with the melody, but I'm not smart enough to do anything other than mirror the bass or guitar anyway 😛 )

[obsidian tendon]

When I started with Suno it was v3 and I thought well if I type in Ooooo ah ah ah it will sing that out. so if I wanted to ring out the word Alone,. Typing out aaaalllooooooonneee would achieve the effect. Wondering if now a days a tag [sound out, echo, etc] would work.

[heady aurora]

Almost every generation it sounds like a robot though, there's no two ways about it it's infuriating

[obsidian tendon]

Question. Should I create the instrumental intro first on it's own using my uploaded audio clip or can I use tags in the lyrics generating to ask for instrument intro before lyrics? It seems like Suno wants to start singing right away:

[Intro Guitar] <doesn't seem to do anything

[Verse]
For you I'd just be there
Someone to share
Talk through the night
Make everything right

[charred merlin]

Hrm -- I know that Suno is a lot like @flint haven's stomach -- what you put into it GREATLY determines what comes out of it...

shiver

Could you share some examples of particularly "robot-y" voices for comparison? I'm happy to blow a huge chunk of my own credits trying to research what-the-hecks and try to figure out how people can workaround v3.5's quirks to appreciate it's benefits ❤️ I already have some results of previous testing over in the pool party @ #1256301881966198805

[charred merlin]

I almost ALWAYS start off with an instrumental intro (sometimes using just one or two [metatags in the lyrics section to tweak a particular style] -- related to the above comment, this adds another "filter" of diversity for Suno to interpret how the lyrics should sound.

This might be why I don't have particularly "same-y" vocals because it's using the existing song as context to infer what the vocals should sound like, instead of just "making a best guess"?

[obsidian tendon]

Got it. I wanted to double check. Thanks!

[charred merlin]

To clarify, that wasn't meant to be "you should yadda yadda", or "you should blah blah", just sharing my own personal workflow to give context to why I might get different results from someone else attempting the same thing, but from a different mindset or approach ❤️

[astral flame]

can i keep a beat and have it sing different new lyrics over it?

[charred merlin]

Not as a "simple" process at the moment.

If you snip out the intro to your song and use the "Upload Audio" feature, if there's enough musical context for Suno to "hook onto", it can continue the beat you uploaded, and then you can extend over it and add whatever lyrics or style prompts you'd like 💪

[astral flame]

or i can remove the vocals and reupload? are these the only options? i have so many good songs with suno terrible vocals and think i found how to get good vocals more often than not. might just have to chalkk those songs as a loss...

[charred merlin]

At the moment there's no way to remove the vocals specifically through Suno ,as they generate the song all-at-once, unlike traditional recordings that have each instrument as their own track when recorded.

https://discord.com/channels/1069381916492562582/1253172964728377455 There's a known suggestion for this idea if you wanted to upvote it for visibility

[astral flame]

on that note, can i buy a bigger plan or do i have to buy more credits

[charred merlin]

But to answer your question, absolutely -- you can remove the vocals and reupload to do the idea I suggested earlier 💪 #📚┃suno-school message Here's a whole suite of BandLab's free web-based audio tools to do every step of this in under 5 minutes without having to even wear pants

[astral flame]

i cant upvote it but can remove vocals myself

[charred merlin]

Sorry for any confusion -- The post that I shared, if you scroll to the top of that link, you can add a "reaction" to the idea to show the developers that it's something you want to see implemented.

[charred merlin]

You can upgrade your account from https://www.suno.com/account

[astral flame]

thank you for the solutions. ripx daw is the best ive found.

[heady aurora]

I would be interested in seeing a generation of yours of acoustic, country, indie pop - something where vocals are important because I don't believe it's as simple as "what you put in" as all the genres I'm trying are poor vocals unless its an autotuned female doing europop

[astral flame]

no worries. only an option for me to view reactions not add. maybe bc i dont use paid reactions on here

[astral flame]

thank you

[charred merlin]

Extra credits don't expire, and are used after your subscription credits run out, but I don't advise purchasing them unless you need -- *make informed decisions about which membership plan works best for you if you don't intend to experiment with it all of the time, consider how you intend to spend your "credit allowance" -- free plans receive 50, every day -- it is slowly doled out.

Paid plans receive all credits at once, plus a bonus

[astral flame]

looks like id have to cancel my plan then buy again? or just buy credits

[charred merlin]

@unique stone , that known-suggestion expired so no one can upvote it anymore 🤦‍♂️

[random brook]

Anyone having a problem where Suno is downloading the wrong song from the iPhone browser in safari or from the app? It's happening in both for me.

[astral flame]

ah. i was wondering why so few votes

[charred merlin]

👉 👈 ... would you be the dearest hamster ever?
-# @astral flame, don't worry I've adopted @unique stone as a community pet against their will -- if they don't want to be forcibly hugged, they shouldn't be so damned helpful and should have picked a less adorable mascot

[charred merlin]

You can purchase extra credits if that makes financial sense to you -- if you already have the Premiere account, there are no consumer-facing options for upgrading, but I believe? you are allowed to purchase multiple subscriptions, if you use another account? @neat nexus maybe could could ask clarification on this matter, whenever the opportunity to address it arises?

[astral flame]

youre so helpful

[charred merlin]

Mama-theharber. is in town, so I'll be out for most of the day

[astral flame]

cheers! be well

[unique stone]

I didn't even realise those threads could expire... The poll in #🗳┃mini-polls is probably the best place to vote for this at the moment since that's probably a more "current" gauge of what's on the devs radar

[neat nexus]

I believe there are no restrictions currently on multiple paid accounts, only on free accounts. That being said if people scroll down, there is a larger enterprise option which has an email where people can contact Suno and request a larger package. Just have to bear in mind that there may be a delay in response.

[young orchid]

How can I avoid too much tearing in dubsteps?

[small sun]

sorry, RTFM... but where do trashed, or thumbs-down tracks go? what if i did it by accident or glitch?

[unique stone]

Thumbs-down is in your library.
Trashed are in the trashcan (icon at top right of library)

[small sun]

lol, i was afraid to click it

[unique stone]

Ah, in case it deleted the whole page? 😄 lol

[small sun]

😱

do thumbs-down tracks go to the end of library or something?

[unique stone]

Should still just be in the same position in the library. They just disappear from the Create page. It's a bit annoying how it does that. You can search for it by name if you remember the name

[small sun]

ok cool. thanks so much. i had a song that somehow turnt into a different one, and the original one disappeard. 🤔

same name, just changed the graphic and renamed it, put it on a playlist, then it was different audio.

thankfully i saved it as mp3 before it decided to become a different composition.

[unique stone]

Huh. Weird. That's gotta be a separate generation. Sounds like the UI fouled up and added the wrong song.

[small sun]

we may never know, but the original is cool and got on the album... maybe some UID glitch or me messing around too much. lol. it's gone, but i have one version that is cool. much appreciated. is there a manual of all SUNO functions? sorry...

checked the mp3 download. which gives a "blob:https://suno.com/" that is apparently unique and does not point to a static file.

i got a file not found page from the original mp3 download

[unique stone]

Things change a lot, there's a "Help" page that covers the basics.

[radiant lake]

Good evening! Has anyone else experienced that during the song the voice disappears from the song without having given a command to Suno? Yesterday out of two versions of a song, I only had one left, the other kept giving an error, although I was listening to it while it was generating and it was beautiful, then in the end it gave me an error and the program withdrew.

[heady sun]

I will just cut the audio short in my video app. I don't really need to do any other edits.

[daring lava]

Please do not post things like this here. No politics @sweet trench

[wet minnow]

no politics please. thanks

[sweet trench]

Suno, when will you have voice cloning available for use in singing in Suno?

[wet minnow]

its already available... called upload

[daring lava]

But you must own the rights to the voice

[sweet trench]

Давид свой чувак? )

[charred merlin]

Apologies, though I believe this user is requesting a feature similar to the voice cloning available through ElevenLabs, if I’m not mistaken?

Audio Upload is currently the only way to use your own voice as the initial upload for a song, but I agree that I would love to see a Suno version of voice cloning 🙂

@sweet trench It sounds like an improved version of this known suggestion, if you wanted to upvote it?

https://discord.com/channels/1069381916492562582/1252789683176345641

[weary heath]

Is it possible to upload Vocals and them to expand without beat? only expand the vocals?

[charred merlin]

Yup!

#📚┃suno-school message

BandLab splitter can help if there’s existing vocals that you want to extract from a Suno song, or you can simply record your voice.

If there’s enough context for Suno to “hook onto”, it’ll be able to continue your prompt, so adding a longer clip is helpful.

It might be helpful to use vocal-suggesting prompts when extending to let it know you want to push it into that style, like “a cappella”

[weary heath]

a thanks. yes u only use my own vocals but actually u can upload only 60s so i want to expand 😅 thanks for help 🙂

[charred merlin]

#1256301881966198805 message

I made a Prompt Experiment Playlist demonstrating vocal-only prompts if it’s any help.

Let me know if you have questions about how to go about something

[charred merlin]

Yeah, if you click “extend”, it can use the 60 seconds for context, and it can continue your vocal tone

[weary heath]

i search prompts to make a remix with my voice for my own song. i dont know if u know the band “HbZ” or “Anstandslos und Durchgeknallt” The remix Voice is changing in the song of them or the newest song of “Die Atzen - Ausnahmezustand” but i dont know if this will possible actually

[neat nexus]

@sweet trench you can probably assume that voice cloning of the type you want will be unlikely since suno is for creating original music. Not for voice clones of existing artists.

[rare oak]

Do you guys use caps in the tags or full min. letters ?

[charred merlin]

I’m not familiar with that particular band, but Suno is capable of some really remarkable things — if you care enough about the idea, and are willing to experiment and try different approaches, I’d be more curious what Suno CAN’T do.

[charred merlin]

Personally I use lowercase, but this is a really fun idea for a Prompt Experiment Playlist when I devote more time to compiling those — i know LLMs can respond differently to capitals, so I wouldn’t be surprised if Suno interpreted them differently.

I’ll try to remember to tag you when I get around to sharing the results ❤️

[crisp fractal]

Hi. I tried to create narrations with sound design and dynamic music in the background synchronized with the narrated content. It works… But the voice is difficult to understand, as it seems to come with low quality and volume. What prompt could be used to improve this quality? Thanks

[supple trail]

My Suno is making the songs with Portuguese from Portugal and I want Brazilian Portuguese. How to change?

The lyrics are my own.

It always worked well. After the changes you made, this problem happened. And even using the Brazilian Portuguese in style of music, the songs continue to come with a Portugal accent

[charred merlin]

While Suno is technically capable of speech synthesis, and I've seen some wild results shared, it's hard to tell exactly how easy those were to make. As well, there always seems to be a quirk to the voice, which I've always attributed to the fact that it's expecting to make vocals or spoken word music -- just my theory.

I wonder if creating the narration with Suno's open-sourced speech-synthesis model? https://github.com/suno-ai/bark That way you could focus specifically on the atmospheric music and add the narration overtop.

I know this is a lot more steps than having Suno do it all at once, but I'd be curious to see what you could come up with 😄

[charred merlin]

Hi @supple trail,

Are you able to elaborate on what changes you were referring to, or when you first started noticing the change in results?

Did the change in results occur between attempts of generating songs using the same prompts, or have you experienced very different results across multiple lyrics/tags/styles?

[supple trail]

Thank you for the answer. The last song I managed to create was on 01/07/2024, but it is in English. After that date, a change occurred at Suno. First the custom started to disable automatically. Every time I log in I have to enable it again. From then on, the songs I want to create based on my lyrics in Portuguese (Brazilian) have come with a Portuguese accent... something that never happened before. I tried several times, including putting the style of music in Brazilian Portuguese. But still, some words come with a Portuguese accent. I need a solution to this problem.

[crisp fractal]

I've done a lot of research and extensively tested prompts in Suno, but it seems to have a mind of its own as the same prompt generates extremely different results. I need to maintain a standard. I must be doing something wrong. But I will still insist, because I liked the emphasis and break in patterns generated. Narrations with Suno are much more expressive and dynamic than 11labs and other neural voices. They would fit well into my project. I will test the GitHub version. Thanks. [edit] I have now tested Bark via Google colab and it is even more unstable. 👀

[crisp fractal]

Yes, I'm also Brazilian and today Suno started mixing his pronunciation with a Portuguese accent from Portugal, which is very different from the Brazilian accent.

[supple trail]

So it is!!! I just want to see how they are going to solve our problem. Because so far nothing. And worse, this didn't exist before 07/01/24. I wrote several songs that came out with Brazilian Portuguese.

[crisp fractal]

eu criei uma bossa nova, mas parece a carmem miranda cantando 😳 era um jingle para um restaurante. Nao consegui reverter.

[safe hazel]

uhhhhhhttps://x.com/suno_ai_/status/1815940718428307605

Suno (@suno_ai_) on X

Available now: Stems 🥳 Pro & Premier users can now separate the vocals and instrumentals from songs, which will give you more control over how to use Suno.

1. Go to Library or Create, and click the vertical “...” on a song row
2. Click “Get Stems”
3. If you’re not there

Twitter · Jul 24, 2024, 2:43 AM

that came out of no where!

[halcyon sparrow]

https://tenor.com/view/if-youre-like-that-gif-13348347855002813843

[real tide]

I just tried it out.
It doesn't seem to separate the Vocals and Instrumental during generation.
Instead, it uses a vocal separation model to separate the generated result.

[charred merlin]

Correct!

This means you can get stems from any song in your library 👀

[royal bear]

It's definitely not perfect. I'm getting bits of each both instrumental and vocal tracks.

It's a start though

[charred merlin]

***Remember, if this isn't your 100% ideal implementation -- #💡┃suggestions-web ***
If you're getting NOTABLY poor results -- #🪲┃bugs-web or ... > Report > Bugs > Poor Audio

https://tenor.com/view/stem-stem-bunnyslide-gif-23881502

[real tide]

Yaa~ 😏
Although I have been using UVR for vocal separation,
the separation results provided by Suno are similar to the results I get with the MDX23C model.
Now I don't have to do this work myself.

[charred merlin]

I KNOW! And to think, this is just what they're still able to squeeze out of the V3x series 😐

[real tide]

Hope they can provide separate Vocals and Instrumental during music generation in the v4

[fluid sentinel]

I can't seem to generate a .wav download for the stems.

I get the thing saying it is generating, just never shows in downloads

Normal songs work

[charred merlin]

I think a lot of people are understandably excited for v4 because it would presumably it would make songs "better" -- we're already getting to a point where "better" is starting to be pretty subjective.

I'm excited to see what new capabilities, if any, the next model might introduce. Kinda like the game Project Zomboid -- they've been working on the next major release for a while, and while people get frustrated and just want the update to drop because "it will be really fun and it looks really good", there's so many backend improvements that it's essentially a brand new game from a modding standpoint.

From a similar viewpoint, while I'm sure v4 will be my most-used model as soon as it drops, I'm more excited about what the devs want to do with the current model, but are limited in / prevented from doing with the current model.

If you're getting NOTABLY poor results -- post the results on #🪲┃bugs-web or report the song using ... > Report > Bugs > Poor Audio

Unfortunately there isn't much technical advice we can give at the moment since the feature is so new, but it IS very important to document as a bug so data can be collected & it can be improved ❤️

[civic frost]

Done. Thanks!

[charred merlin]

Thank you, dude. Having technically-minded feedback about this with results to compare is super helpful. 🙏 Can't wait to see how this gets refined 🙂

[sage flower]

Glad to see stems get implemented. Hopefully it’ll lead to a more useful feature being implemented on the backend such as reusing the same vocalists

[cunning rock]

This may be a dumb question, but can I ask how suno references the songs to generate? Sometimes I’ll explicitly state in custom I want an ancient sounding Middle Eastern song and the closest it will generate is a modern Celtic style song

[patent leaf]

What’s a workflow for dealing with noise in v3.5

[crisp fractal]

I didn't understand if steems conversions (separation of vocals and instrumentals) are charged or not. Because I separated about ten songs and I was deducted 30 credits. [edit] (he cashed it and then returned it) 👍

[patent leaf]

Genuinely riveting advice. You are a genius!

“Do what makes it not happen” you should be a professor

[fluid sentinel]

What type of "noise"? Distortion? Aliasing? White noise? Hum? Etc

[steady granite]

Has anyone yet found a way to .. go back to edit a clip, changing .. say.. a couple of wrong words, and then regenrate the same clip intact with just the new vocals being different?

[untold briar]

I have occasionally posted long bits of strategies for fighting noise and sometimes even winning, or recovering partially at least. But it was more of an intellectual puzzle. Practically speaking, if you spend the same time and credits to start start over and make new songs from Clip 1, you probably end up with a higher quality song.

That said, if you don't mind changing your song (the voice, the music style, or even both) you have more options. Basically follow the advice for a 'genre shift' where you use a series of very short clips (no music history) but instead of style shifting to a new genre, you style shift to better audio quality. Then after the audio is better Get Whole Song to put it back together again. But you almost always end up with the music genre or the voice or both at least drifting and avoiding it entirely is a lot of work. But if you happen to be making a song where you don't mind change, then this can be a good strat.

There are also some things you can try using Upload to re-create a song, not exactly, but close enough sometimes. But you have to do audio editing yourself so you end up with pieces of a full song - not a full song on Suno.com, if that matters.

Fingers cross for some future prompting/song-history features to make this more of a winnable fight.