📚┃suno-school

[charred merlin]

ugh -- the automoderator bot is hyper sensitive. I'll ping someone so they can repost your message -- a lot harder to fix on mobile 😅
(that ugh is at Woodhouse, not you 🙏 )

[cinder slate]

Anyway I was trying to say it’s weird how few parameters there are basically

[charred merlin]

@unique stone, if you're around are you able to save the last woodhouse'd post?

[cinder slate]

Like I would expect a stable mature model to be able to do things like “play just a woodblock once every quarter note” but this is like impossible AFAIK

[charred merlin]

you rock, hamster

[charred merlin]

You absolutely can get that -- but not through "commanding" it.

If you were jamming with a musician who doesn't speak your language, (perhaps they're from a non-verbal culture, and are JUST learning words, so they don't understand ANY of the archetypes that you do), if they misinterpreted "woodblock once every quarter note", you have a couple different options;

1. try to learn their language and say it EXACTLY with zero room for ambiguity (might be ridiculously hard)
2. repeat the exact same thing you just tried verbatim, albeit louder or more slowly (condescending, and assumes that your collaborative partner is an idiot)
3. show them exactly what you mean using the available tools you've both agreed upon (using audio upload to show it 6-60s of woodblock every quarter note so it can understand and follow along)

[cinder slate]

That’s good to know, I will keep experimenting, if you have trivial examples like this I would love to hear them

But like if you make an analogy with the coding models, you don’t get good results by saying “make me a fun app that my kid would like” you go through it piece by piece and this is not my experience working with suno

[charred merlin]

I'd say it's not SO far off -- because you can get really interesting one-shot generations from AI trained on making websites -- like https://websim.ai/, or these quick websites that I asked Grok 2 to "no-code" for me

That in mind, the "one-shot generations" from Suno are AMAZING, and often are top-40 radio quality in and of themselves, but as I say regularly, the more specific of the song you hear "in your mind's ear", the less likely you'll be to achieve it as a one-shot generation through the use of complex prompting -- you'll need to break it down into multiple segments with extensions and create it -- the exact same way you would approach building an app with an LLM -- start off with the foundations, UNDERSTAND AND TRY TO BREAK THOSE, then slowly add things to the stable foundation otherwise it's all a house of cards

lmao, little sneak peak of the stuff I google search

[cinder slate]

Well ok true some of the no-code tools can do stuff like that… but it’s clear that you can “talk to copilot” about code and it “understands it”. My ideal music-gen would understand like “ok 30s intro swing 80bpm the chord progression is E major A major, then… etc.” like it doesn’t seem like it “understands” music if that makes sense, maybe just me

[charred merlin]

Sounds to me like you've spent a fair amount of time learning the quirks of natural language processing LLMs, and have a reasonable familiarity with what you can expect as output from them -- would you say that's correct?

[cinder slate]

I’m a software engineer so yeah I have worked more with chat bots and coding but also midjourney and stuff. Maybe there is a lot about suno prompt engineering that I need to learn (I learned a bit) but usually I get the best results by just rolling the dice again and again (and the results are really awesome that’s why I’m here don’t get me wrong)

I can also produce music so I can take the output and finesse it but I’m taking about the technology itself, the API seems really black-box and strange(?), even compared to Udio IMO (not sure if we’re allowed to talk about them lol)

[safe hazel]

Suno doesnt like to reveal a a lot about their model

probably related to the lawsuit

[cinder slate]

👆

This was my cynical reasoning about it

[charred merlin]

Or just related to the fact that it's a trade secret, and there's literally no financial benefit to sharing them.

[safe hazel]

Yeah I'm sure the lawsuit has nothing to do with it at all

[charred merlin]

Doesn't seem like they need to be mutually exclusive, especially how the lawsuit is attempting to get them to reveal their trade secrets. But I feel like arguing the same opinion from different sides of what I hope is the same page.

Coca Cola sure doesn't like to talk about the content of Merchandise 7x, and KFC doesn't really like to talk about their herbs and spices, regardless of whether or not there's lawsuits, because that's what makes those companies those companies.

If they revealed so much about their methods that anyone else could just copy the exact same thing, Coca Cola wouldn't be as standout of a brand.

[cinder slate]

Like are there any white papers about prompt engineering for text-to-music? I haven’t found any. But prompt engineering GPTs there are like real scientific papers it’s not a secret how the foundational models work

[charred merlin]

I agree that there's a big gap between the way we'd LIKE to use Suno vs the way that we currently interact with it, but keep your eyes out for developments in the industry, because this isn't just a "Suno" thing.

LANDR is developing a new set of training data to enable companies to access something akin to a song-side API for greater control over prompting. https://www.landr.com/fairai/

[desert geode]

It appears that Suno creates a unique version of the voice by utilizing its tone and pitch rather than replicating the voice exactly as it is. To me, this seems like an intelligent approach to generating a variety of voice types.

[charred merlin]

But I urge you to think back to the days before you were familiar with LLMs, back when it was purely "magic in a box" that you didn't understand the twists and turns for -- it's still magic to you nowadays, but you've grown with it and learned to use it collaboratively in a way that neither of you could have done as well in isolation.

It'll be the same with yourself and Suno sun-enough

[charred merlin]

I might have misinterpreted the original question, but extending from a voice isn't the same as using it in training data. The extension simply analyzes the patterns in the waveforms (probably oversimplistic analogy) and then tries to continue.

Training data would imply that after uploading a segment you can recall it from a BRAND new generation that doesn't extend off the first. I've seen nothing suggesting this.

[cinder slate]

The proprietary chatbots are still based on public research and published scientific ideas, and you can control them in very fine grained ways. I hope this is the case for Suno et al. but that’s not my impression

[desert geode]

That was the case. At least something really similar to what generated after extending from my voice uploaded.

I can send you the examples when I have time

[cinder slate]

https://arxiv.org/abs/2306.05284 maybe i just need to dig into this stuff more and see what the open source models can do...

does anyone know/use any good text-to-music models on huggingface, e.g.? would love to know

[charred merlin]

Just to clarify, did you 1) extend from an audio upload using your voice?
or
2) make a brand new "create song" that WASN'T an extension, where you heard your voice?

If it's the former -- that's what it's supposed to do, so that it can try to continue the music you provided it -- this is NOT the same as longterm vocal training, or anything similar to ElevenLabs vocal cloning etc. Any patterns regarding your particular vocal timbre, etc, aren't retained and trained upon, they're only considered at runtime when "Extending"

[desert geode]

The example I sent was not the best, but I think I’ve deleted the others

[unique stone]

The closest thing you'll find relating to how Suno works is Bark (open-source text-to-speech which I think uses the same underlying technology.) Both Suno and Udio don't make public any of their inner workings and although Udio for example gives you a little more control over things, there's no public information at all as to the internals of what those controls do.

Most likely the reason they don't make this info public is because they're running a business and it's essentially "trade secrets", rather than hiding info because there's something naughty going on. I would love for them to open-source their older models as new ones are made available (e.g. Chirp v1 isn't available at all any more, it would be nice to let people see how that worked.) But, at the same time, some other company could start up as a competitor. There's also a lack of documentation due to no prompt being 100% guaranteed to work and it seems like prompts are also learned as part of the model rather than being input manually during training.

[charred merlin]

https://stability.ai/stable-audio I haven't looked into it, but the same team behind Stable Diffusion released Stable Audio Open, alongside their closed-source models

[desert geode]

I sent you the example where I extended from my voice, and then the song I didn’t extend from any upload. It wasn’t the best examples and I need some more time to find the better fit to the previous generated voices.

[cinder slate]

"trade secret" is one thing but if it ends up making the API/app really obscure and hard to control then it's kind of weird, I mean MidJourney has plenty of secret sauce but their API/app also offers a lot more options than the music-AI apps

or ChatGPT, def tons of stuff going on that we're not privy to but the API is pretty sophisticated at the same time, anyway that's kinda my thoughts, interested to see what yall thought

[safe hazel]

"Any patterns regarding your particular vocal timbre, etc, aren't retained and trained upon, they're only considered at runtime when "Extending"" do you actually know this or are you assuming?

[charred merlin]

🙏 ❤️ ☕

[charred merlin]

#📚┃suno-school message
I'm happy to engage in grounded speculation, but I try to base what I say confidently based on observations and what little information is made available to the testers to help them give informed feedback.

[unique stone]

It seems to me that what Suno actually does with uploaded voices (or any sound really, even instruments) is just try to mimic it based on what the model was originally trained on. So it'll never be 1:1 the same but it'll be pretty close.

[charred merlin]

As a guide you're a bit closer to "how the sausage is made" -- do you have any insight to whether or not Audio Upload is actively used for RLHF in the current models?

[cinder slate]

slight tangent, are we even allowed to try to understand how Suno works? this is an interesting clause from the TOS:

...You agree to not use the service to:...

obtain or attempt to access or otherwise obtain any content or information through any means not intentionally made available or provided for through the Service;

[charred merlin]

Armchair lawyer -- if you use any privilaged information (like an ex-employee using stuff that wasn't made public), you're in trouble.

"don't hack, and you're fine"
since it says "intentionally made available", if their source code was leaked, you're still not allowed legally to copy it and make a competitor

[unique stone]

The devs keep any inner workings like this private, even from me. The most I would have insight into is potential upcoming features (as part of alpha testing, which you also know of.) As far as I am aware the models themselves remain static after release, and don't get trained on any user-uploaded content. There seems to be some layer in-between the prompt input and the audio generation that they could tweak manually (there was that bug where songs kept saying "start" at the beginning for example), maybe some parameters that control the general audio rendering behaviour, and possibly some auxiliary functionality such as the gluing process involved in "Get Whole Song" that can be adjusted. All of these seem to be manual processes though.

[cinder slate]

does "hacking" include like "prompt hacking" for the sake of understanding the model? this is not usually against TOS for other AI stuff

[charred merlin]

Prompt hacking is NOT hacking

If Suno didn't want you to try to learn, they would have banned furrybeans, paimon and I a long time ago.

As long as you're entering text into the lyrics field without the intention of injecting malicious code, it's not hacking. You're being experimentative.

[cinder slate]

lol ok like i said i'm new to the channel, sorry if i'm talking too much, i'm only here cuz I think the app is awesome, but I'm just asking some questions that have bothered me

[cinder slate]

doesn't this border on "reverse engineering" maybe(?)

[unique stone]

It's healthy to be curious about how all of this stuff works. When I first found this it was a real eye-opener 😄

[charred merlin]

Perfect place to ask them, please don't consider any shortness or terseness as tension toward you for asking questions ❤️

[cinder slate]

no not all 🙂 thanks

[charred merlin]

As long as you're using the service they way the intended (by entering lyrics or style prompts) you're in a safe playground.

Reverse engineering is what the scam API sites do -- THAT'S unintended use of the service

https://suno.com/song/197cffc1-da6e-4fb6-afe7-362b62f0c84b
If Furrybeans is allowed to do this in the lyrics field without being publicly chastised by the team, I'd say you're safe to "Break" the site by creative use of lyrics and styles

(yes. This will always be a point of reference now, hamster <3)

[cinder slate]

lolll ok cool yeah this is the kind of stuff I'm interested in haha

[charred merlin]

https://suno.com/@furrybeans https://suno.com/@sixstringtoker
"Experimental" playlists from established Sunovators is a great place to look at how to "creatively break" Suno -- actually experiments and iterative learning instead of "rolling the dice" with a boatload of metatags and style prompts until a song respects enough of them to trigger your confirmation bias.

These beautiful weirdos actively try to break it, and then once they find something weird, they prod at the wound until you get something legitimately horrifying, or oddly beautiful.

[safe hazel]

This isn't an answer

[charred merlin]

#📚┃suno-school message

If the way we've phrased things didn't help in the slightest, then I'm not sure we'll be able to give you an answer you'd be satisfied with. Speculate however you'd care to 🙏

[safe hazel]

I'm not the one speculating, you are. The answer is: it's an assumption. You can just say that

[atomic iris]

Having great difficulty getting the program to actually generate in the right style of music - I tried to create a polka track and it gave me straight up heavy metal. - Appears the genre of music might be heavily weighted on the lyric content?

[hollow hare]

3.5 is known to be more focused on particular styles of music. For genres it cannot seem to get right, I highly recommend that you start an intro piece (favorably as an instrumental) in 3.0 — which is better at different genre styles — then Extend it as 3.5 for the better song structure

[ancient patio]

Hello , how can I contact support?

Today I've been trying to pay for my subscription and it said my card was declined , subscription was not activated but my card was charged.
where can I open ticket

[charred merlin]

[e-mail] for account related issues. 🙏

[ancient patio]

thanks mate

[atomic iris]

Thank you

[dull arrow]

Is it possible to add a feature that lets you control how similar we want the song to sound to our uploaded audio? Like a 1-10 slider

[flat galleon]

how to adjust the voice on this track to sound more soulful and with a greater vocal range? I'm aiming for an R&B vibe rather than the current K-pop or Disney princess style. I want the voice to have more depth and emotion, characteristic of classic R&B music.

[unborn wigeon]

How can I help get my language into the Suno database?

[fluid oyster]

tried a new song in your language is it ok ? https://suno.com/song/58757dd9-3a91-49d4-b1e2-cc356a9c523d ? thx

I mean pronunciation mistakes and so on

[unborn wigeon]

Suno will write the lyrics in my language but the pronunciations are off. Is there a way I can help improve the pronunciations of the words? Input in the correct word pronunciations into the database?

[whole comet]

Is song structure the Suno equivalent of hands for Midjourney? 😄

Having real problems with structure, loopbacks on choruses, etc. - even with the "simple" interface and having Suno generate the whole song. It won't even follow its own lyric structure.

Done some digging; tried using meta-tags, tried using some alternate workflows that worked better than the tags ( the Markdown version in particular ), but still had issues. Dug around in search engines and articles. Checked out Reddit. Nothing suggested seems to work all the time.

Is the only way to explicitly guarantee a song structure ( especially alternative ones like Choruses where each iteration has slightly different lyrics ) - other than getting lucky - to extend and input each section separately? Or am I missing something obvious ( which is entirely possible 😄 )?

[safe hazel]

the repetition of verses is a widespread issue that I don't think anyone really has a fix for. From my own experiences when I went from always structuring every verse as 4 lines to using different numbers of lines for a verse the repetitions happened less often (but they still happen sometimes). So it seems to me like verse structure has some impact on repeating verses but I cant really go much further than that.

[whole comet]

Cool, thanks. Yeah, I suspected that was just a system issue that we'll have to wait for it be developed/trained/evolved past.

I haven't actually tried extending and adding each verse/chorus/bridge/instrumental separately - but I've heard that works. A lot more work/credits - but if it gives the structure one wants, it could be worth it.

I'll give that a try.

[frozen pivot]

The less tags one uses within the song, the better chance for non repetitions. (At least for me). The focus should be in style section of the song.

However, some repetitions are classic… no problem in repeating a strong verse

[arctic cedar]

I don't know how to use discord very well. Sorry if I do all this wrong. Im trying to find out if anyone else is having problems with the suno player from their profile pages. My bar is not working today but always has before. --- nevermind I think I fixed it on my end .. haha browser tab was muted.

[sleek canopy]

suno error go brrrr

[true night]

they added small improvements iin suno i notive i could not expand this before now i can

4 hour here we come?

[true night]

quite intresting indeed

[true night]

we all been there , im more like were the f is the music coming from 😄

i came to relise i added the project thing in gen chat was a large reach. im prob better off adding it in here or in offtopic even though it is quite suno related. but i will come back to that when i rewrote it a bit to look more eh. proffesionale :)

[astral flame]

i was using a prompt only no extention or upload. its happening again, short songs. maybecause i listened before the song was done being created. it never did that before but seems to make short songs if i listen before its done now.. maybe. thats all i can think of.

[split vapor]

Just like before the release of V3.5, the sound quality is poor and the music is very short. There is also little variety in the music.

My credits are disappearing in vain

[celest fulcrum]

hey everyone, can someone help me with a billling problem?

[shut gulch]

Three weeks ago, the music generated was of very high quality. Once I had chosen the genre and vocals, everything fit together and the results were what I expected. Unfortunately, over the last week the funds I spent have been wasted. The quality of the music generated has dropped dramatically and the results are significantly different from what they used to be. What happened? Why has there been such a decline in quality?

If I spent 30,000 credits in less than a month, did I get moved to another lower quality plan?

[charred merlin]

V3.5 excels at song structure, but can be “generic” to start songs off.

Try using v3 to start your song, and extend using v3.5 to “solidify” your amazing intros and have better lyrical flow 💪

[clever bloom]

How do people collaborate with each other on here?

[sterile yacht]

Someone starts off with lyrics and another person adds to it .. , the original user extends the song with those newer lyrics > rinse repeat.

So one user will be the generator and another the extender of the lyrics.

[clever bloom]

But how do they make the actual song? Does one person download it from Suno and the other upload it?

[sterile yacht]

after all parts have been generated they just use get whole song, and publish the final output

[violet tundra]

it's a process of sharing links... the extensions are made from the link shared.

[clever bloom]

Gotcha

[lime mauve]

Hi, I use SUNO today to generate a Chinese song, when I put the lyric words” 红鞋”means red shoes it sang in Cantonese. I changed the words to”红色鞋” which is the same meaning. It sang correctly in Chinese. Is this because the dataset was not clear? I do know there are classical songs in Cantonese often use word”红鞋”.

[graceful condor]

Specify in the prompt "Mandarin Chinese" and see if that works. Write the prompt in English and the lyrics in Mandarin.

[polar spade]

Going off on a wild hair. It could be interpreting any language into " Chinese" , first of all, so, if you write in Mandarin, French, English, whatever, it interprets in to Cantonese by default. But that assumes Cantonese is it's default "Chinese" language. Keep in mind that Suno is based, as I understand, out of San Francisco, an area prominent in Cantonese people, if chatgpt is accurate. So there may be that influence to choose which language is "Chinese". I wonder what music and songs suno had used to learn from, and which songs were available, he thinks rhetorically.

As wuv says, try Mandarin, instead of "Chinese".

[lime mauve]

I’ll give it a shot

[untold briar]

I don't know of a simple word you can add to the style to force Suno to use Mandarin or Chinese. My guess is a simple solution probably exists. In other languages, users have found using text formats that are common for children or language learners sometimes help. Diacritics, or latin accents in languages that would typically use capital letters to indicate stressed syllables.

One prompt method that works for language accents is to add a lot of very obvious accent phrases. For British English "Could you do me a favour and pass the biscuits?" or something silly. You can add these after your real lyrics, and then Extend the song from before these extra phrases so they are not part of the final work. They are just used to make your Song look 'more british' to Suno. It's possible something like this would work for Chinese or Cantonese. Maybe adding very obvious phrases could do the same thing.

[whole comet]

random repeats, dropping lines, and interjecting new ones - this is "excelling at song structure"? Structure is the one part giving me grief on this platform.

I'm sure it has improved from the beginning. I'm also sure it will improve over time. It's still a PITA now.

[flint yacht]

Same

[charred merlin]

I know you're dissatisfied with repeating verses, etc, which are common "hallucinations" that we're willing to help with here in tech-chat, but if I can explain the difference that I mean in "song structure" and "lyrical flow"?

When you would expect 4 bars for verse, it wasn't uncommon for v2 and v3 to attempt to fit 6 instead of 4, or would introduce weird timings as if it had borrowed or lent a beat earlier in the song -- weird timing stuff if you're familiar with music theory, or just "funky parts" of a song otherwise. 😛

v3.5 is almost mind-reading level when it comes to adhering to a logical flow for how the lyrics should be sung (this is separate from lyrics being repeated or dropped due to investigable "hallucinations"), and it absolutely blows v3 out of the water in terms of being able to "hook onto" an existing pattern and continue from it (so much in fact that you're unable to extend audio uploads with v3 or v2 because they aren't powerful enough to give consistent results 👀 )

Good to keep in mind that none of this is "magic" yet, and the biggest determining factor on Suno's output is your input, but if you have any specific questions, we're happy to help step by step 😄

[remote rivet]

I'm getting a false-positive content match on this audio I just recorded. Can anyone provide a solution?

[charred merlin]

Oh no wait, it IS literal science fiction magic.

[charred merlin]

You might need to contact [e-mail] and send them the file you've recorded to help them improve the autodetection system 🙏 Unfortunately there isn't much we can do from the discord-side of things 😅

[frozen pivot]

Great explanation 🎉

[charred merlin]

[safe hazel]

My favorite is when it will cut off mid-WORD to repeat the previous 4 lines. It's weird that it seems to repeat 4 lines, not a discrete verse section, but if it starts repeating after the second line in a verse it will go back to the second line of the previous verse (when I use 4 lines for every verse)

[frozen pivot]

Yeah!
It certainly does that. Like in a real music studio, artists do mixup their lines sometimes. That’s why there is “Punch in and Punch out” 😊

Suno has “Extend” from a certain time. Looking for more improvements in this area

[whole comet]

Like I said - I'm sure it will improve over time - and I'm having fun with it "as is". I'm sure you're aware that other AI projects have similar "quirks" - look at image generators and hands 😄 They eventually got it; I'm sure music generators will eventually overcome their unique hurdles.

I'm less interested in whether Suno is perfect - it isn't and it's not realistic for me to expect it to be perfect - but whether there's workaround for structural idiosyncrasies.

[regal haven]

Hi I just started using suno today and I got a question? If I got like a small snippet of a song, then can it convert it into a full song? It's an existing piece of art though?

[charred merlin]

I agree that technology improves over time, but you raise a perfect comparison between image-gen and music-gen 😄

NOW we're at a point where models like Flux can generate text and fine anatomy like hands without any complex prompting or "know-how", but it was always possible if you were willing to go deeper than a "one-shot generation".

Sure, in image-gen there were "textual inversion" collections like 'bad-hands' to use in the negative prompting field (https://discord.com/channels/1069381916492562582/1253341197917884530 fingers crossed we see it in Suno one day! 🤞), but those were still basic adjustments to help guide "one-shot generations", and could only help nudge what was going to happen anyway.

For those who have specific songs or images in their mind, there are always people poking, prodding & starting bleeding-edge discussions in the github comments -- that's sort of what tech-chat is like. We try to figure out how people can use the existing tools to get the songs that they want.

He'll argue the fact that I don't have a crush on him, but @naive quail is one heck of a Sunovator, and all of his "official" songs likely, conservatively, have 5+ extensions each in order to take "a good one-shot" and turn it into "the fully mastered song that he hears in his head", or at least taking steps toward that.

Suno absolutely isn't perfect, it's trained to create on autopilot given little-to-no guidance, and learning how to make the best use of it is an ongoing learning process for new users, advanced users, and devs alike 👀

And then as soon as you think you've mastered it, they release a new model that acts ENTIRELY differently, with it's own strengths, weakenesses, and best use-cases when combined with others #🪲┃bugs-web message (for instance, v3.5 goes "generic" without strong prompting or a hook to continue from, but can still give interesting results, while v3 is more broad and versatile as the model dropdown suggests, and I've just recently learning that v2 ISN'T included as a novelty, and legitimately has powerful use cases 👀 )

[charred merlin]

If it's someone else's song, Suno is designed to stop you from using it. 🙏

[regal haven]

I see 😭

[charred merlin]

[naive quail]

My instrumentals average ~ 10 to 12 extensions each.

For tracks with vocals, if I have a something specific in mind, I'll go the extension route and usually do an extension for each part of the song, like I do for the instrumentals (Intro, Verses, Choruses, Bridge, Solo, etc...). Suno will occasionally churn out a 1-shot gem when experimenting with lyrics and styles to see how Suno responds, I'll happily slap a cover on those and call them done when I like 'em enough.

[charred merlin]

I completely understand the passion behind the idea, and it's not a matter of saying personally "you are not allowed", but just working within "fair use" law.

Perhaps you could record a fan version of it, and while you still technically wouldn't have 100% ownership of that, it's at least a pro-active step to handling things in the right way, seeing as how there MIGHT be a grey area in what's allowed from region to region -- but if you have any serious intentions, it's always best to read the https://suno.com/terms terms of service, or get "official" advice from [e-mail] or a copyright lawyer 🙏 💪

I am excited to hear your song, however you are able to create 😄

[whole comet]

Thanks - that was pretty much what I'd concluded was needed if you wanted a specific and/or unorthodox song structure.

[regal haven]

I can use a cover and it could work?

[charred merlin]

Is this a cover that YOU YOURSELF have recorded? Or still someone else's work?

[regal haven]

I mean I can record one but yea I was talking about another cover

[charred merlin]

To be fair, that's the exact same situation as uploading someone else's work, except now it's starting to seem like you're trying to find ways around giving credit to the original artist, or putting "creative transformative effort" into it.

I'm just an armchair-advocate, and my approval or disapproval really doesn't matter, but while I'll always advocate for people choosing how to approach a problem, IMHO, there actually IS a right and wrong way to handle using someone else's content, so I'd advise that you contact a copyright lawyer to discuss exactly what you'd like to do with someone else's song, and what legal rights you have to do so.

Like, if you wanna learn guitar and play "Hey Jude" at a campfire, Apple Records isn't gonna crash your birthday and request a licensing fee. But this is a privately held company, making clear communications as to what they're allowing their product to be used for. Again, I'm not the dude to say yes or no, just trying to point you in the right direction to figure out how to responsibly handle the matter if you love someone else's song and want to embrace that "ownership" that you feel over it -- there's powerful and proper ways to do so ❤️

[regal haven]

got it 👍

[charred merlin]

Since we usually rap about this kinda stuff in private messages, I'll pivot here for a quick second if it could help others -- I know that you still put a LOT of work even into your one-shots, but that breakdown of the extensions shows a lot of passion for a specific song.

Since I think a lot of people "turn off" when they see these kinds of suggestions because some users may believe they "don't have enough credits to learn that kinda way" or they "shouldn't have to pay to actually learn how to use Suno".

I know personally how I feel, but do you have any free time to share any insights on your different types of songs?

btw, I just realized the ▷ symbol on the playlists that are intended to be played beginning to end like albums. I friggen love you dude.

[naive quail]

The method works well and allows you to extensively tweak and nitpick the hell out of each section (as long as you have credits to burn). I just wish we could have Suno generate smaller extensions for a lower credit amount - I often only use 15 to 30s from any extension and I'll typically know in the first 15 seconds, or so whether I'm going to save the extension. Hate having to pay the full amount for each of those, especially when I discard tracks before they've even generated 30s of audio.

[naive quail]

Maybe some day 😄

[charred merlin]

I'll bribe you with deathwish coffee and whatever else you're into -- I dunno, foot pictures or something.

[naive quail]

hahah

ooooh... coffee beans between the furry beans!?

[charred merlin]

Suffice to say, the songs with multiple extensions probably mean a LOT more to you though, right?

on average?

[naive quail]

😄

Still don't feel a strong ownership, but yea... those take more effort.

[whole comet]

I think that'd be a great enhancement: credits per second, with selectable ( or at least boundable ) seconds. Still - I can't do partial images on MidJourney, and I've iterated literally thousands of images in "mid-stages" there - so it's not just Suno 🙂

[naive quail]

Even just breaking it up in 15s billable blocks would be good... and have it stop generating the second we toss a track into the trash - that's soooooo much wasted computing power.

[whole comet]

And honestly - with AI tools - even the time and cost of iterations are way less than trying to build it by hand 😄

[charred merlin]

I know you've not come around to that, in my opinion it may be because you're a LITERAL multi-instrumentalist and haven't felt like you've needed to put a personally satisfying amount of artistic expression into, or the learning process was "too simple" to feel responsible.

Either way, I always find it fascinating what random instrument you've decided to focus on for a week when you get bored, seeing what Suno can do when you make an instrument out of a garden chiminea or something

[naive quail]

or 26.5 GHz standard gain horns 😄

[charred merlin]

Or wtf the orcs blew at Minas Tirith -- terrified to ask where you get these things, btw

[naive quail]

and the Udu is NOT a garden chiminea, damn it!

Kevin blowing his nose... lol

[charred merlin]

I mean, your vases aren't technically vases, but they still work the same? (there's a joke in there somewhere about using a bong as a vase, but I'm not smart enough to make it)

[naive quail]

grumbles

One even holds water.

[violet tundra]

The answer is "no." In order to use anyone else's song, you must have a license. If it's a cover -- the people who covered the song have the license, not you. If you use the song to create a song, it's fairly easy with today's tech to see what you've borrowed -- even if it seems different to your ear.

[regal haven]

alright-

[charred merlin]

I'm honestly surprised how easy it's becoming -- while you still need to front the costs yourself, a lot of distribution partners from DistroKid to LANDR are taking the heavy lifting out of setting that up -- you let them know the song, pay the money ,and they try to secure the rights 👀 --literally EVERYTHING about the music industry is changing, not just music production. Slimebag managers are trying to rebrand themselves as talent partners and whatever because people can just do the same stuff, better, faster, and often for next-to-free ❤️

[violet tundra]

Not sure I follow. the distribution places are making it easier to obtain performance rights from BMI and ASCAP etc.? (Deleted content b/c I'd have to do quick research)

[charred merlin]

I'd have to look up the specific integrations, but if you let them know all of the licenses you're applying for, and can fill out the forms that your distribution partner provides you, they work with companies like the Harry Fox Agency to legally acquire the rights to use the song commercially

It's a value-added service though, so simply being partnered with DistroKid or LANDR doesn't immediately mean that you can go wild with it, but it means that they're willing to front the leg-work so you can release your music through their service, instead of people simply not being able to legally release covers because they're worried about legal ramifications.

[alpine venture]

https://suno.com/song/56d88e94-e422-4af6-a7d3-16723a01f0b7

[flint haven]

This is all very interesting but ultimately pointless, as it has already been pointed out that at this time it is a TOS violation to do covers.

[alpine venture]

Ends with a literal full minute of silence

Kind of fun

[charred merlin]

Just to wrap things back to Suno-specific though, even if you reach out to the Beatles and get permission to remix Hey Jude, as noted in the popup and in the TOS, you need to have EXCLUSIVE rights to it, meaning that you've managed to convince the Beatles to let you be the only person in the world who is allowed to use it. This also explains why public domain is a POWERFUL legal concept for how much creative transformation needs to occur before you can claim any sort of copyright, but that's an entirely conversation altogether.

Wow. thanks woodhouse. That's kinda rude, we're discussing the nuances of the matter and trying to help people make informed decisions, but I'll try to keep things on topic.

[flint haven]

Rude? You yourself posted an image of the TOS regarding uploads. I was confirming what you posted. But never mind.

takes a shot of prune juice

[charred merlin]

It was a nuanced discussion about why this type of behaviour isn't allowed through Suno itself, as they're a private company and set their own ToS, however, if someone is interested in exploring how to LEGALLY produce covers, I'm trying to encourage that conversation in general.

I see that #📚┃suno-school isn't the appropriate place for these types of talks though.

[safe hazel]

did woodhouse just become sentient

[charred merlin]

I apologize, @flint haven, you're absolutely right. 🙏

[violet tundra]

Wow. Go Bot!

[exotic narwhal]

Question: How to can I improve quality of Suno generated audio in terms of artifacts (sound distortions)?

Results are amazing but it’s easy to hear there are frequently random sound distortions and other artifacts. Any well-known approach to fix these?

[flint yacht]

Not sure what's going on

[eager topaz]

The only sure way I have found is to use RipX DAW Pro, rip to stems and use the cleanup features available. RipX have a 21 day free trial if you wan to try it, but it's the best I've found

[south leaf]

Is 320kb/s coming soon?

[split vapor]

I tried that too, but the sound quality is bad in V3 and I don't like the music either. I guess I'll have to wait until V4 comes out.

Do you plan to update the version? It just keeps consuming credits and is difficult to use as it is now.

[torpid mortar]

Tried to distrubute a song i made with my suno pro plan, and i got this message: Rejection Notes
Please make sure to fix any errors before submitting:

You appear to be distributing generative AI content. Please note that most AI content is currently not eligible for distribution, as is outlined in our Terms of Service. Please only resubmit with a license that fully outlines terms of use. Thank you!

I have full rights to use songs i made with my pro plan? And is there somewhere i gan get license that confirms that?

[astral bay]

the platform u wish to distribute dont allow ai simple as that

[whole comet]

Doesn't sound like a problem with your "right to distribute" - it sounds like the platform you're trying to publish on doesn't allow generative content.

[torpid mortar]

But the platform says they accept ai aslong as its mone😂

Mine*

[astral bay]

well they state that "most" so theres some tiny % that is allowed but not yours as it seems

[whole comet]

https://sunoaiwiki.com/tips/2024-05-05-how-to-resolve-copyright-claims-on-suno-ai-generated-music/

[hollow hare]

Which platform is it? I think SoundCloud had something similar stated where they do not let AI generated content in unless it was made with one of their partners (Fadr, Soundful, and Voice-Swap)? I have to look back into that

https://press.soundcloud.com/234082-soundcloud-partners-with-fadr-soundful-and-voice-swap-to-introduce-artist-first-ai-integrations

[torpid mortar]

I use soundcloud 🫣🫣

[torpid mortar]

I added screenshoots if all that a few days ago, they still declines it😰

[hollow hare]

Hm I double checked SC general TOS and it did not mention declining AI work unless the content was an impersonating work.
https://m.soundcloud.com/terms-of-use

Though their monetization program do explicitly reject AI work:
Covers, remixes, mixes, mashups, DJ sets, audio recordings (e.g., spoken word works), podcasts, and recordings that are exclusively computer-generated through the use of artificial intelligence or any other means or method are not eligible at this time – do not submit them. https://artists.soundcloud.com/terms

[torpid mortar]

Any suggestions on who i should use instead? Example Distrokid?

[hollow hare]

I heard DistroKid is okay for beginners until it isn't. I know Robert said he was burnt by them.

Also DistroKid services do not keep your songs up when you are no longer subscribed unless you opted into a leave a legacy plan (which is much cheaper to do as an album)

When lookong into distributors, you MUST look into at least two items in the TOS:

• How the royalty is paid to you
• What happens if your account ever gets terminated.

Even if you think the 2nd one won't apply, you know what they say: hope for the best, prepare for the worst. It can be as little as someone submitting an unsubstantiated claim or you enlisting into Google Ads to get your account cancelled

(the latter due to some distributors thinking you got botters with the amount of traffic the Ads can bring)

When you have time to spare, check out Top Music Attorney TOS coverage (links at #🗣️┃chill-chat message)

Sure her videos can be click-baity but the TOS really helped me with some of the legal terms used in there.

She also pointed out how suspicious TuneCore and Symphonic TOS are because they state the company will retain your royalty if your account is terminated for whatever reason, which could be as little as a baseless user complaint

[south leaf]

I used Ditto to distribute my music, they have a 30 day trial.

Ditto keeps your music up, even when you stop paying.

[tight spade]

thats a gpt

such as https://chatgpt.com/g/g-i4alc1CRd-suno-ai-helper

or theres a few hundred others

discord was built for gaming

suno in my opinion makes more money off people try diff stuff then giving out a bible that would streamline the proces

trial and error makes the money go around

with that said im not oposed to it

[grave wraith]

I wonder if it wouldn't be possible to think about a feature to produce your prompt instant as stems? The stems feature is great, but as we all know you can't really seperate voice from music 100% clean as the frequencies merge too much.
TBH I don't know if that would be even possible for the AI to create the song not mixing instantly

[unique stone]

At the moment it's not possible as the AI is trained on fully mixed music so that's all it can output It'd need to be trained on stems which are probably a lot harder for the team to get hold of. Would be very nice though.

[open field]

Is there an overview or a list of tags that can be used in Custom mode? I often have the problem that songs are sung in such a way that there are hardly any breathing pauses, which makes it difficult to perform such a song with my own band. Is there a tag for inserting short pauses?

[charred merlin]

Unfortunately there isn’t a place to post this kind of suggestions, since the discord feedback channel was shuttered.

#🚧┃suggestions-layout might be an arguable fit, but just sharing the idea so that your suggestion can get more relevant eyes on it for longer ❤️

The devs don’t poke their heads into tech-chat for ideas too often — this is the laboratory, and maybe the devs get scared that we’ll pull up a chair and ask them to stay and answer 100 questions 😛

Either way, I love seeing that people care — not discouraging the post, just trying to find the most effective place for it 😄

[desert geode]

Moved it here 👍 #🚧┃suggestions-layout message

[charred merlin]

💪 sweet, but I didn't mean to make you feel like you needed to delete the original post 😅

I wanted to make sure the idea didn't lost in chat or end up missed by the devs 💙

[pseudo sphinx]

/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

[alpine venture]

https://suno.com/song/b58e6fd9-085e-4d9a-ab47-57135944dc0b

@untold briar thought you might find this opening hallucination interesting

Since it's between the words

Oh waaait, its from the prompts

I didn't see the word up there

Or the style rather

[frozen gulch]

Ehi but the creation of songs is only on the site or there is also a bot here?

[unique stone]

It's only on the website now. There used to be a bot a few months ago but it was a very old version of the AI.

[obtuse ridge]

Can someone fix this song, I cannot get Suno to sing the first verse going like "I was born in the USA..." https://suno.com/song/73581fa4-b085-4694-945c-265d0ffc7123 Here is the Audio snippet used to make it: https://suno.com/song/cf7feb33-8ef5-4515-bdf2-8a60999cdbb6

Also the piano stuff after the Outro can go, be cut away

[turbid nest]

i got the subscription just for the STEM option, but the thing only separates the vocal from the rest..That it?

[flint haven]

That is correct.

[timid wasp]

And yet the voice is heard in instrumentals.

[turbid nest]

yes, its quite bad...I can use 3rd party software to do the separation, but need a ton of work to clean in and i always end up doing everything the old way (making everything on DAW). Just cancelled the subscription because its kinda pointless to me if i cant have proper stems. did devs ever said anything about actually making it possible?

[silver kraken]

Anyone have any idea when the system maintenance will be done? Do they have a system status page or anything like that where they elaborate on what the maintenance window will be?

[blazing heath]

Clean stems and removal of Sibilance and artifacting in vocals and drum cymbals would be an amazing feat!

[austere dragon]

And the vocals sound incomplete without the vocal "ghosts" that show up in the instrumental stem.

[timid wasp]

Yup, very confusing story.

Sometimes I wonder doesn't it generate tracks too fast, too hurrying. I wish there would be something like "master mode" where it takes longer but you can get every track from a song clean.

Lemme dream 😭

[rose jasper]

Do keep in mind, that if you add your vocals into ( ) prefixes, these will also be included into the "Instrumental" STEM

[turbid nest]

yep. it´s a ok pastime to people who doesnt know music, but a bit boring for music producers... some extra option like tempo, music progression, time signature would be nice too... i try to tell the AI but not sucessfully... lol

[grim flax]

I have a question, not sure if this is the correct place for it. I want to have backing vocals, all different kinds in my tracks, especially backing "Ahhhs" maybe ohhhs" , not sure how to go about this. Any thoughts or ideas would bge very much appreciated.

[turbid nest]

nah, the thing i would like is just separate tracks so i could remix...i dont like most of the tracks made, but some parts can be nice. So, in the end. i gave up separating them, so in the end i just listen to it and play myself at the DAW. But damn, it could speed up a lot things with separated tracks 😄

[quasi summit]

You can try (ahhh) for background vocals or Ahhh [Talking]. Or AHHH... to emphasize and extend.

[neat nexus]

You can somewhat get this with (ahhhh) or (ohhh) but the AI will take off on you and start doing it on its own after a bit.

[rose jasper]

If you want to use SUNO for STEM seperation, you'll be sorely disappointed, it's not it's main work flow.

[gusty pendant]

How long will the update be going ?

[neat nexus]

And this is absolutely the right place for this discussion.

[turbid nest]

try:
(backing vocals) Aaaaahhhhhh
(choirs or watever) XXXXXX

[turbid nest]

yeah, already realised that...just got annoyed that i subscribed becaus of the "STEM" option... kinda misleading... 10€ is nothing for me, but a lot of 10€ from everyone is a lot of money....

and does Suno makes anything other thasn 4/4 time signature? i asked 3/4 and 7/4 but only game 4/4! lol

[ashen juniper]

Moises is much better than suno stem seperating

[turbid nest]

hahahahaha

https://tenor.com/view/big-bang-theory-sheldon-cooper-bazinga-ballpit-gif-3558458

[ancient patio]

are u guys planning on making official api?

[unique stone]

It doesn't understand numbers (time signatures, tempo etc.). "waltz" in the style tag has a chance to make it generate 3/4.

[turbid nest]

and they say AI is the future..

but i guess its fun to get kids into music

[unique stone]

It takes a while to get to the future 😄 I'm just enjoying the journey

[turbid nest]

the way things are now, the future is more like "Idiocracy" movie.. 😄 i think artists are safe 😄

ok, to be fair with AI, the music is as good as modern pop music/spotify top 100

and i mean "good"

[turbid nest]

maybe using melodyne to detect everything and make it a .midi file could work for me.... uhm... why i didnt thought about it?!?!?

[balmy rampart]

Can someone help me install the api?

[neat nexus]

Can you link the API for us?

[charred merlin]

Hey @balmy rampart,

There might be a degree of confusion I'd like to try to clarify -- are you following instructions that you've found elsewhere, looking for help to figure them out?

Or are you simply asking if there's an API made available?

[tacit thorn]

Any ideas as to what causes the lyric generation errors?

Or how to resolve them?

I just got one for a prompt that should not have any issues

[true night]

any suggestions how to not make a song that im packing into a song not crash and go away 😄 (waited 30 min 1st try) i hope i dont have to sit and try this for 6 hours lke last time 🙂 with 100's of "failed to generated clip\song"

not that i can really do anything beside pray that suno do finish it so i can see how long it has become.

[tacit thorn]

Ok, I resolved my issue. Interesting that the hangup was providing a name to a described character

[sage sapphire]

Is suno genertae songs in kids voice

[flint haven]

✅ Warning logged for iamrafi55. I couldn't DM them. || Please do not post the same question in every channel. Give people time to respond.

[frozen pivot]

Interesting!

So you have a DAW, and you know music. Here is my suggestion. For now, as Suno works on most of what you are looking for (We all want the same or similar features); Use Suno as your starting point for your professional music production. This includes generating sound ideas, melodies, vocals, etc. Take the stereo track to your DAW and do STEM Separation. Since you are a professional music producer, I will assume you know about Steinberg's Spectral layers or Izotope's RX. Both can go a long way in restoring or separating audio tracks into stems. I am sure you would thank Suno for what it is today.

Here is the link to Spectral layers
https://www.steinberg.net/spectralayers/

Here is the link to Izotope's RX
https://www.izotope.com/en/products/rx.html

[true night]

okay i need help rendering a song ive tried for 3 hours now (3 times) to get the full song and i get an error every time after around 1 hour (50-1 hour) ever time.

[balmy rampart]

Is there a Suno API for me to create my own application?

[frozen pivot]

Not yet

[balmy rampart]

If I want to sell music automatically using your service, is it not available yet?

my idea was to create a suno child panel

[frozen pivot]

Try to login with a different browser or clear your current browser cache to eliminate that aspect

[balmy rampart]

carpa help me, please

I don't want to generate music at a lower price. I want my clients to pay me to generate music and for me to make a profit.

[true night]

guess its not getting fixed then ive tried that on 4 browseres 2 that was brand new (i dont use other browsers . ) thank for sugesstion. (i should prob mention its a 3 hour (almost) song. )

just havbe to keep create full song for days , it worked after 1 week of trying last time about a month and half since

[frozen pivot]

😀 It's not our music. it is yours. You can do anything you want with it. However, make sure you are a paid subscriber to own the commercial license for distribution

[balmy rampart]

Yes I know, but how do I have my own application where my clients can create music and pay me?

I managed to do this using systems for other ias. but not yet for sun.

[frozen pivot]

With the API, which is not currently available

[devout osprey]

Who is the support team lead?

[frozen pivot]

Not here. This is a self service group to help each other. For Suno support, send an email to Suno.

[balmy rampart]

Isn't there an extra official API so we can automate our sales?

[devout osprey]

that makes sense now. ok ty.

[unique stone]

Suno do not currently provide an API.

[balmy rampart]

ok thank bro

[frozen pivot]

Trust. We all are looking forward to that. Hope soon

[frozen pivot]

👍🏻

[balmy rampart]

You can make infinite sales when they release this function

[frozen pivot]

I agree

[graceful condor]

where do i find the key

[unique stone]

This is the wrong discord server. We keep getting people joining asking for a key. This is an AI music generation service.

[timid gazelle]

I have an odd question and a request...
First of all, 80% of my generations don't work with [Female voice] and [Male Voice] prompts, it's either random, or it's just one voice.
Is there any documentation on what prompts in brackets can be used so you could add more depth to the song?
I know the * for echoes, and (for ad-libs) and [Piano Solo]
But I want to know what else I could use, since there's little to no documentation found.

And now the request.
It would be nice to have a song "seed" or "base" in case you like the way it sounds, and you would like to use the same generation but different lyrics, or same lyrics and voice but different soundtrack.
Or a way to generate/inspire a generation via your previous songs (not reuse prompt feature)

Or even a way to download the vocals separate from the soundtrack, and vice-versa.

Also, for a way to alter/edit the same song.
For example, I didn't like how my verse 2 sounds, and I'd like to change the lyrics just there, instead of generating another song.

[flint haven]

@graceful condor https://tenor.com/view/ghostbusters-close-shut-door-bye-gif-15786234

[frozen pivot]

Within the style section, "female-voice" or "male-voice" mostly work for me. Seed is one of the most requested feature. Yes! It would be a great idea to have the same voice in different tracks. Let's see what the future holds

[timid gazelle]

I would like to add different instruments in a song and even add choir, but the choir feature didn't work for me, the piano solo and instrumental prompts mostly worked for me, now i'd like to add for a different song organ solo.
Has anybody tried that?
What other prompts works?
Like mention a key the voice should be in, like a deep bass, a growl, a scream, etc.?

idk if i'm allowed to link a song posted on YT here (being made by me using Suno) but i'd like to make something similar, and expand upon the generation, and make more vids for my channel

[polar spade]

I just finished a "cubano" (latino dance) style song that I wanted to get some good harmonies in the chorus, so I've had luck on it and others with adding "barbershop" without my song sounding like a barbershop quartet , but caveat that cubano has a strong barbershop harmonies anyway. "A cappella" can work on occasion as well. Suno pulls the harmony concepts (frequently along with a bass line) of a cappella as long as its later in the prompts and not always just "no music". After all, a cappella is Italian for "as in the chapel" or.... choir.

[supple silo]

So is anyone else getting some kind of 403 error?? Iv haven’t been able to make as song in a couple weeks now..
I’ve tried different browsers I’ve tried clearing cache cookies extensions the only thing that works is signing innon a different account

[timid gazelle]

I didn't mean the auto generated song via "insert what your song type erc.." I mean the custom mode

[timid gazelle]

Works good for me, did you check if you have credits?

if you do, and it works on a different account, try opening a ticket

[frozen pivot]

[Instrumental break: organ, deep bass, growl, scream]
Just try and see. Who knows. Also add "organ, deep bass, growl, scream" in your style section

[timid gazelle]

I'll try once my free credits drop today 🙂

[lunar anvil]

growl and scream for a choir?

[timid gazelle]

no, choir at the chorus for example, and the ad-libs to be scream, like for example

[Instrumental]


*(Verse 1: Male)*  
In the valley of echoes, where dreams fade to rust,  (*dust*) <growling
I stumble through shadows, in this world of dust.  (*rust*) <growling
With every step I take, I plunge deeper in the mire,  (*wire*) <scream
Regret wraps around me like a noose of barbed wire.  (*mire*)'' <scream ```

[lunar anvil]

Gotcha. That makes more sense.

[tranquil dune]

wow

[timid gazelle]

and I wanted to replace the piano solo with organ solo, and add after the verse for example, bagpipe solo, just to see how it sounds in a metal/rock song 😄

[lunar anvil]

I tried adding a choir and never could get it to trigger.

[timid gazelle]

same here.. i got around it by making the whole chorus as ad-libs via (insert the text here)

for example:

{Fingers grasping at midnight, gasping for air.)  (*despair*)
(Together we rise, though the darkness may fall,)  (*CALL!!*)
(In the shadows of the summit, we hear destiny’s call.)  (*FALL!!*)```

it's a "simulated" choir

[lunar anvil]

I will have to try that.

[tranquil dune]

i'm trying to make a duet but it's not working

[timid gazelle]

that's what I was saying 😄

[tranquil dune]

😭

[timid gazelle]

tried that for a week, and after ~100 songs, i've got 3 that had remotely worked

[lunar anvil]

I get duets when not trying to get one.

[timid gazelle]

odd

if it's against the technical aspects of this channel, I apologise, i'll post the link of my song here, to point out the way i kinda wanted the song generation to look like.
It's a more "realised" song that kinda worked, but not really as intended, but it sounded good.
https://www.youtube.com/watch?v=bZtuyUo9GOk

[polar spade]

Got it. Ignore that. Reverse it.

[timid gazelle]

the song's backbone is this:

*(Verse 1: Narrator 1)*  
In the darkness, I recall, (call)  
Whispers soft, they start to call. (recall)  
Memories dance in the night,  
Echoes fade, but dreams are bright.  

[break]
[bass drop]
[instrumental]

*(Verse 2: Narrator 2)*  
I remember when we laughed,  
Under stars, our secret path.  
Every glance, a spark ignites,  
Together we can touch the light.  

*(Chorus: Both)*  
[choir]
Come on, let’s chase the dawn,  
Feel the whispers, hear our song.  
In the night, we rise so high,  
With your hand, I’ll touch the light.  

*(Verse 3: Narrator 1)*  
Underneath that silver moon,  
Life was sweet, we found our tune.  
You were there, a fire bright,  
Guiding me through all the night.  

*(Verse 4: Narrator 2)*  
Echoes of our laughter ring,  
In my heart, you are my spring.  
Through the shadows, we will fight,  
Hold me close, we’ll touch the light.  

*(Chorus: Both)*  
[choir]
Come on, let’s chase the dawn,  
Feel the whispers, hear our song.  
In the night, we rise so high,  
With your hand, I’ll touch the light.  

*(Bridge: Narrator 1)*  
Let’s remember all our dreams,  
Every sigh and silent scream.  
In the darkness, love will guide,  
Together, side by side.  

*(Bridge: Narrator 2)*  
Every moment, every breath,  
With you here, I feel no death.  
We’re the echoes of delight,  
Together, let’s touch the light.  

*(Chorus: Both)* 
[choir] 
Come on, let’s chase the dawn,  
Feel the whispers, hear our song.  
In the night, we rise so high,  
With your hand, I’ll touch the light.  

*(Outro: Both)*  
[choir]
In the darkness, we’re alive,  
Whispers soft, our hearts will thrive.  
Hand in hand, we’ll shine so bright,  
Forever, we will touch the light.

[Short instrumental]
[fade out]```

[polar spade]

Does your genre typically have duets in the real world? I do notice that "romantic " and "kpop" will do better to sneak in duets on my songs. Obviously if a banger hardcore screamo song, you may want to try the prompts towards the end.

I'll post a stupid romantic song to demonstrate.

Kpop though tends to start with female vocals anyway, if you add a rap, it will switch to male often enough. A few generation , say 3 out of 7 I've gotten duets. Again, depending on your direction and style, it's trial and error.

[tranquil dune]

my genre is mostly romantic

[lunar anvil]

I use the (phrase) styles all the time. I have been trying to get the choir harmony, but I only get something when not trying for it.

[polar spade]

Well, stupid song here. But I thought I had romantic in it. Oh well. It still did Duet. Likely because suno recognized it as romantic to begin with.

https://suno.com/song/abb833d7-28f9-4d19-8bc1-3a60bcf5d1b8

Oops... I hate the photo, sorry. How to turn that part off?

[tranquil dune]

ooh okay

[naive quail]

Suno definitely recognizes "Choir" in the style and as a tag in the lyrics, but it may be one of those things that gets easily overpowered by other styles/tags to the point that Suno just ignores it. It may help to over-prompt for the choir or actually start the song with it, which might make it easier to pull it in elsewhere in the song structure.

[true night]

ok now i tried everything ive tried 4 diffrent browsers , cleaned the cacke coockies relogged restarted ect. my get full song still fails i even to to extend it and take an eariler part no matter what i do get full song fails. what to do?

[naive quail]

just did a quickie (and silly) choir test: https://suno.com/song/d7dffbe5-0141-4909-9aac-2ef075d65e68/

[timid gazelle]

contact Suno via a ticket

[true night]

meaning* email?

[tranquil dune]

https://suno.com/song/85f9fbb6-6a6e-4c66-85db-b75c80c4945b

[timid gazelle]

there should be a e-mail or something, idk if discord has ticketing system or report an issue section

[polar spade]

I've had luck with, and depending on your taste, with "gospel" and "barbershop" . Sometimes choir does work as I think many use it, but I'm not sure what suno pulls out of choir style that is NOT harmonies, but I've had more luck with the mentioned two.

[naive quail]

The lyrics can definitely play a big part in whether Suno does a duet or not.

[tranquil dune]

i make songs out of my poems

well try

[polar spade]

I'm at a restaurant now. So I can't listen at the moment. Sorry. I'll add you and catch it later.

[timid gazelle]

would be nice to see the prompt list with [prompt] to see what Suno baked in as prompts, etc..

there's a limited list, and in howto's there's little info

[polar spade]

I see your prompts. At least try [Verse 1: male], [Verse 2: female] ... others have tried [Verse 1 #male] or Verse 1:#male] and so on...

[timid gazelle]

tried, it's 80% hit or miss

[naive quail]

I'd bet that there aren't many baked in other than those related to song structure

[tranquil dune]

Gotcha

[timid gazelle]

even in the auto generated area

[true night]

yh they have an email thing but i dont think they can help with it but i can try.

[timid gazelle]

try the website e-mail, I guess it's account related

[naive quail]

[e-mail] ... it may take a few days for them to get back to you, tho.

[polar spade]

I haven't tried but I'll try "antiphonal" next time too. That's a choral style where each phrase is song by another singer. That should be interesting to how suno interprets.

[true night]

ive know it usualy take about 1-2 weeks to reply . i was just hoping someone knew a magic trick to get longer song to well not go error

[lunar anvil]

I've gotten duet inside the same line, every other phrase, and every other verse.

[polar spade]

Don't forget, "a cappella" and "broadway" ar style that incorporate solo parts....

Oh and BTW, definitely try [alternating vocals] in the style.rather than "duet" If suno recognizes it, I think it's helpful. It worked for me twice.

*are styles that incorporate......

[polar spade]

Yea. Peppering the song seems a reasonable means of trigger suno to get the idea. ⚡️

[tranquil dune]

I don't know why it's nbot letting me create the song D:)

[polar spade]

( I can't type and eat at the same time apparently)

[lunar anvil]

I tried adding duet in the style. Did not get a duet so I removed it and the the very next song with a different style created a duet. I think the AI was trolling me.

[lunar anvil]

Have you tried changing the version from 3.5 to 3 and back again?

[tranquil dune]

no 😭

[tranquil dune]

it still didn't work

[lunar anvil]

Are you out of credits?

Or have you tried getting the full song on a different device like your mobile?

[polar spade]

Did someone mention harmonies a while ago...? I forgot, also, [ CCM ] (contemporary christian music) , known for its emotive grandiose builds can sometimes give some harmies on chorus to your song with changing your style.... much.

*harmonies

Harmies?lol

[safe hazel]

Use asterisks makes echoes? Is that an agreed upon thing?

[polar spade]

So I hear. I use ( ) which work so I havet tried **s

[naive quail]

Not really. paratheses have worked best for me (for me) (me) (me...)

[tranquil dune]

do i actually have to pay for credits D:)

[lunar anvil]

I use ()[echo] to get echoes of echoes.

[timid gazelle]

Yes, or if you are on the free plan, you get 50/day

[safe hazel]

@polar spade @naive quail I see, thank you!

[naive quail]

here's an example: https://suno.com/song/13a0d143-7eb8-4896-9849-346282b5539c (singing) (singing)

[safe hazel]

oh that's cool

so here the (* sigh *) gave you the effect. Is that how you do sound effects? Not with [brackets]?

[tranquil dune]

i love this

[tranquil dune]

makes sense

[safe hazel]

credits refresh at 12:00 am UTC

you can make more songs then

[naive quail]

That's one way, at least. There aren't many that work consistently (except maybe for (*beep*) and they don't always give you an effect right where you drop it... or even always the effect that you want. Definitely fun to play with, tho!

[tranquil dune]

question do y'all post your songs to youtube?

[safe hazel]

gotta take all these notes ✍️

[naive quail]

taking it a bit darker... this one resulted less in an actual echo and more of a... second darker voice. https://suno.com/song/135504dd-0dc6-4ec0-9d75-dbbdf0118622

[safe hazel]

holy....

[tranquil dune]

omg

[naive quail]

[tranquil dune]

crazy

[naive quail]

I did an entire dark poem with various (*sfx*) after each section. I cleaned them all out afterwards... my secret recipe https://suno.com/song/173ac910-bc00-4c67-8834-ca70eac7de79

(*wind*) was one near the beginning. It rarely works, but seems to have in this case.

[safe hazel]

it seems like nothing works every time. And all we can really do is add some biases to what we want to happen

[naive quail]

Suno has a lot of sound effects in the training data... we just don't have a reliable way to call on most of them 😦

I've had a car engine start and even a horse whiney. gunshot is fairly reliable

[safe hazel]

i wish they would give us some kind of library of available effects to try

[naive quail]

At this point, we likely know what works better than the devs do. 😄 It's AI - I bet they're surprised by half the stuff it does.

[safe hazel]

yeah that's probably true, i wonder how often they test it out for themselves

[lunar anvil]

If they ever publish an API we may find out the secret sauce.

[prisma wave]

Anyone know why nearly all forms of rock are turning into country now? Really sucks

[lunar anvil]

Have you tried changing the style to something else and then back to rock?

I find that sometimes after many generations of the same style it goes down a path it wants. Just by changing the style to something else then back to the original it is almost a like a reset.

[polar spade]

On suno or the real world? Way too many hiphoppers doing country now. I had to double check my spotify radio to see what I was streaming.

[prisma wave]

I'm talking about Suno

[naive quail]

the lyrics in v3.5 can have a pretty strong effect on the style, so if they match up with something in the training data that's associated with country, the track may end up leaning country. Might need to over-prompt some rock related styles/tags

...and maybe remove any references to beer, pickup trucks and your dog.

[timid gazelle]

yes, I do.
https://www.youtube.com/watch?v=bZtuyUo9GOk

[tranquil dune]

I subscribed

[timid gazelle]

Thank you!
I intend to post every day
Already I have the schedule to post a video every day till Sept. 3'rd
That's all the songs that I've made that were "good"
I'll probably use the other generations in the future, but I don't want to post them so close together

Is the "EXTEND" feature in Suno not working?
I'm trying to add to an already generated song, but it cut out at the end.
This is the song:
https://suno.com/song/2f5c9a60-ac66-4bf0-b9c8-4dad3faa48f2

[tranquil dune]

i do want to post my songs aswell

[timid gazelle]

Nvm... I've figured the "Extend" feature out...
Had to extend it 2 more times..
Got some editing to do...

Audacity FTW 😄

[hollow hare]

In case you were unaware, you can combine the extended parts by clicking "Get Whole Song" on the latest part you want to combine (option appears in the menu AFTER that part generates).

It is recommended to combine the first extension then extend THAT new Get Whole Song for consistency (though there are reasons you may not want to do so, such as force changing the music style with the termination extension method).

If you also were not aware, check out https://suno.wiki

[rustic wagon]

Question is there any prompts that can reduce the artifacting in the song generation?

[safe hazel]

You can try something like "Clear Male Vocals." I always use "Clear" to describe the vocals, it seems to help but, nothing works all of the time

[timid gazelle]

Thanks, I made it!
https://suno.com/song/44d0b4f2-d689-4517-970e-1fc3b2c8281d

[frozen pivot]

I use clear-male-voice or clear-female-voice all the time

….with good results mostly

[potent rock]

if i had a song that i exteneded, and i liked which was part 2, but then i exteneded that bc part 2 wasn't long enough, how do i make it a whole song? I can't combine part 2 and 3

[loud inlet]

i have this exact same doubt

[calm bone]

I've been getting 403 error for a few days now

[dry gale]

If I have monetization on my YouTube channel, can I use music created by Suno AI?

[leaden lark]

you can only monetize the songs you make on the suno plan, any songs made before that are sunos and can be used non commerically

[dry gale]

I have purchased a subscription. Is it necessary to mention Suno's name in the description when I use the songs?

[leaden lark]

did you make the song while subbed?

[dry gale]

yess

[leaden lark]

they you should be good if you made it on the sub, the song is yours to do what you want

[dry gale]

Thanks

[violet tundra]

"Get Whole Song" will combine parts 2 and 3. Once you get the whole song, since you like parts 2 and 3 (3 added because 2 wasn't originally long enough) you can then extend again - and again, etc. Ultimately, you can then lop off part 1 in a DAW or video editing program and retain parts 2 through 10, 11, 12, and so on.Once you're done extending, get whole song again and you're done. Note ... extending without occasionally uniting them via GWS, might morph the music style... Extening from GWW will retain the overall music style.

[spark palm]

https://suno.com/song/51a28f98-31aa-43e8-a9d7-f211c1e551de

[frozen pivot]

Yep!
Get Whole Song from the three dots is the way to combine multiple extensions 👍🏾

[languid jolt]

Good day. I generated a bunch of songs last night to try out new styles and so on. Wanted to download the songs today, but they're gona. ALL my previously generated songs are gone. Anyone know why this happened? Is this a thing with Suno?

*gone

[unique stone]

Log out and log in again, making sure you have chosen the correct account. It's pretty easy to select the wrong login option 😄 I have done this before. Hopefully it is just this and nothing more serious.

[near flume]

Some on YouTube and some on Bandcamp.

[cunning mulch]

So, uh... wanna hear a cursed out-take?

Have fun: https://suno.com/song/d69c0c43-fa39-4d17-bc31-2f9114f92842

[hollow sapphire]

https://suno.com/song/157b13a7-1481-4403-a3f6-e2f7db4a6847
😎 Hello! The song I'm presenting here comes from a text that has been sitting in a drawer for about ten years. Originally, the text was in the form of dialogues between a mom, a dad, a grandma, a grandpa, a little boy, and his sister! However, Suno doesn't yet allow for this type of exercise with multiple voices; at least, I haven't been able to manage it. At most, two voices are possible! So, I reworked the text to adapt it to the mom's voice... I really hope you enjoy it...

Ultimately, I am rather satisfied with the result, although a hint of frustration remains from having altered a text that I was particularly fond of, especially since it is, of course, the kind of text that is marked by personal experience... So, I still want to know if any of you have encountered this kind of barrier during the creative process?

[tired needle]

why i still get 50 credits every day even though i have pro plan

[naive quail]

#📢┃announcements message

[rich timber]

🤚 hi, new here

Is there anyone available to assist me?

Is there anyway to correct recording or use the same music and singer?

FunSoundDesigners678

[shut moat]

query: what would you describe grimdark music as?

[real falcon]

Hi all,
Is there a way to lower the music / increase vocal?
I often find myself in need of downloading the stems and edit it in another music App to change the volumes

[shut moat]

im writing a death march chant in Khazdul.

dwarven.

the lyric gen doesn't know what khazdul is but it seems to be able to speak it.

I want to make it more...fantastical.

like the start of the song is kind of what I want the lyrics to sound like but then it just whispers the rest of the way

https://suno.com/song/1c52ed1c-885d-4b4a-8f7b-3da7266c5624

[rich timber]

Is there anyway to correct recording or use the same music and singer?

[shut moat]

i thought you did that with extensions.

im a noob don't listen to me.

[real falcon]

fantasy, dark rock?

[rich timber]

How can I get someone from Suno to help me?

[shut moat]

i don't want guitars.

i want like a marching chant and war drums and maybe a little bit of grandiose ambient symphony.

you ever listen to grim dark music?

that + Doom hell priest chant + powerful operatic shouting.

https://suno.com/song/8e827c52-9e37-4011-b459-0f667411ea6e

it's getting closer.

[real falcon]

fantasy can be done in many styles

[shut moat]

well you can do fantasy with your vision then

I appreciate the input, but what I'm hearing in my head has no electric instruments.

[naive quail]

The only official help is via email through [e-mail] . This discord is a user community - any help you get here will be from other Suno users just like you.

[real falcon]

you're correct, dark rock would deffo have those instruments.
Fantasy/sci-fi could perhaps hit that style more.
Though I think perhaps it would fit better with the style/mood rather than a genre

[shut moat]

hmmm i wander if acoustic organ metal would work.

nope.

Grimdark ambience chanting, tribal grimdark marching, industrial marching, slow grimdark hell chant

this got me the closest to my vision.

you just have to hammer it in sometimes.

[cold zenith]

is there a problem with the website ?

[violet tundra]

"Correct recording" this could be a lot of things .. a few specifics would help me answer. Regarding getting the same voice -- once you find an iteration with the voice you like, simply "extend" from a point in the song, making sure to keep the original voice within the captured part. If you want the same voice and stanza beat, make sure the extension as both voice and stanza. Same thing with voice, stanza, and chorus.

[rich timber]

Do you have videos for training to use the system?

[violet tundra]

By correct recording, meaning something is off with the instrumentation or the way the vocalist sings your lyrics - unfortunately, the only way is to do a reiteration. Just make sure the extension has the "correct" way captured at least once, then after a few iterations, the vocals will likely come out the same (or very similar).

[rich timber]

I remember seeing "extend". I just didn't know how or when to use it

[violet tundra]

You're asking me his as if I'm "Suno." I'm not. I'm just a schmuck. You might find some training on YouTube -- I have never checked.

[violet tundra]

Oh, wow. You're really new.... okay..

Let me login..

[rich timber]

I'm old, but I'm new 😆

[violet tundra]

FUKHING WOODhouse stole my explanation ... TOO LonG FU WH.

[violet tundra]

Okay, Discord's bot is not allowing more than one-sentence replies.

Login.

Look at the menu to the left.

Create, library, explore, etc.

Click create.

Fu WoodHOUse

once you click create... you'll see a window to the left.

(beneath song description)

Enter your preliminary song instructions here. [alt-rock, percussion, singer-songwriter, what have you, separated by commas]

[about: asteroids crashing down in a city of people in love.] Then click Create

You'll see your song being made. to the right of your window... when the song is ready, you'll see two icons with song "title" in bold white and a red dot. The dot indicates you have not listened to that track.

Play the song... the song will appear in a song player at the bottom of your screen. If you're like me, you have the font turned up a little... so zoom out to about 75% or 67% see the song player at the bottom.

In the song player will be three dots. If you want to build on the song sample, you can click the three dots and extend the song. Or download it.

"Extending"

If you like your song sample, you can extend the song by clicking extend. A new custom window will open, allowing you to enter long lyrics, the time stamp from which to extend from (00:14 or 01:24, whatever, enter the time you want to cut off from. Everything before the time stamp will stay in the next iteration. Everything after will be removed from the next iteration.) Enter lyrics, Click create. Keep going, more or less, in this fashion until you get a song you like.

That should get you going.

[stray slate]

Hey y'all, I have a short question that I thought Stems might be able to do the job but not.
I generated a few songs that I like the vocals in it, so I want to create more songs (same style) with the same voice. Is this technically possible? Like having your own "vocalist"

[rich timber]

^^^^ I know how he feels 😆

[violet tundra]

typically by just extending with the approved vocals in the song. Not really from song to song. But, the same "singer" will show up again, if you have similar lyrics and the stame style set. From my experience, Suno has 20 to 30 voices. More or less.

[stray slate]

I'd say in Spanish it have less, but we play with the accents

[rich timber]

Do ya'll create on a computer? I've been just using my phone (ducking head in shame 🫠)

[violet tundra]

I'm actually the odd one out on that... I do not use my phone for anything. (not sure how the UI looks on the phone).

[rich timber]

It looks small 😕 😆

[violet tundra]

Sorry, I gave up wasting credits trying to get accents.... Can't help you there..

[violet tundra]

LOL

I can replicate a phone... gimme a sec...

[naive quail]

I took one look at the interface on the phone and immediately NOPE'd.

[violet tundra]

The create button is at the bottom of your screen.

click create, and a new window rises up from the bottom.

[rich timber]

Ok. I will go to my computer later and follow directions to the Mars section in cyber space

[violet tundra]

Enter in song description [new-age mystical piano] [about song subject, etc]

okay

[rich timber]

I appreciate your help. I have a feeling that I will be coming back to visit again soon 😆

[violet tundra]

Glad to help!

[umbral sluice]

Why are ALL my songs coming out with female vocals now. EVERY. SINGLE. ONE.

Even when specifying Male voice/male vocals/sung by man

It's ruining the whole experience of writing lyrics. Especially when you have a specific voice, sound, or vibe in mind.

[violet tundra]

I've said it before -- Suno is 1) sentient 2) passive aggressive.

[umbral sluice]

Sounds like an excuse. I want to know WHY.

[violet tundra]

yowza, you are wound up. I wasn't offering an excuse. I was making a joke. To temper the tone. (the joke offered in lieu of my not knowing why Suno sometimes gets on petty kicks).

[umbral sluice]

When I write lyrics I write with a specific perspective and tone. I don't need Suno deciding my lyrics should be sung in any voice but the voice I speccified. It's as simple as that.

[violet tundra]

Yes, try this..

Suno is likely calling off of some old vocals....

[umbral sluice]

If that's the case I am clearly wasting my money on Suno

[sharp solstice]

Can I determine type of voice by metatag??
I'm trying to make duet by men and women, but I don't know how to do that.,. 😦

[umbral sluice]

that's because you apparently CANNOT do that. Suno decides regardless of your input and instruction.

[violet tundra]

You're not wasting your money-trust me. You're having a difficult 10 minutes. Yes [Male Vocals] put this in the song description and in the song style box. First, however, refresh the window. Repaste your lyrics into the lyrics box. Reset the time stamp 1 second earlier or after. This will get rid of most earlier influences. Try again.

[umbral sluice]

Incorrect I have already done all that. Still get female vocals on EVERY song I generate.

[violet tundra]

Okay, that's fine, something in the lyrics can trigger a femal vocalist or a male vocalist.

[sharp solstice]

do I have to type [male vocals] in style box too??

[violet tundra]

Okay, it's just a matter of trouble-shooting. Keeping everything the same, enter in a new set of lyrics. Same female vocals? If not, your lyrics are triggering something.

[sharp solstice]

Suno originally has to repeat it many times to get good results, but I think it's really hard to make a voice of the gender you want at the desired time...

[violet tundra]

Also, the [song style] will impact the vocals. You can adjust the style description without really impacting the output style. For instance, [k pop] can often lead to female vocals. But you can change this to [pop with percussive synth] and get something similar sounding but primarily male vocals.

[violet tundra]

I would.

[sharp solstice]

Then if I wanna make it duet male and female, what I have to type in style box??

[violet tundra]

And yes, trying to get a duet triggered by timestamp--huge credit sink.

[regal zinc]

why were my credits I paid for deleted? I still had over 400 credits and they were just deleted

[violet tundra]

Frankly, everyone wants a duet. I did. I couldn't get the change how I wanted -- by stanza -- so I relaxed my demands and took the duet I was given -- stanza / chorus. Duets are something Suno does, but not a feature as such. You're in experimental mode, and it leads to frustration and loss of credits.

[violet tundra]

This happened to me... Not knowing your situation... mine just reappeared. I refreshed, and there they were. SO .. a UI glitch. Not sure. 500 credits. That sounds like free mode. To have them disappear -- for good. never heard of that.

[stray slate]

A friend of mine mastered it, he is the one inspiring me to get the same voice in different songs, I think you trick to just extend will work somehow

[regal zinc]

Ive refreshed.. it still says I have 0 credits left..

[violet tundra]

"A friend of mine mastered it" is he a good enough friend that will tell you if asked?

[regal zinc]

why is there no place to talk to support?

[violet tundra]

😕

[violet tundra]

How many songs did you create?

[regal zinc]

I literally had 415 credits left an hr ago and now its says I have 0. I made only 20 songs

[violet tundra]

Okay -- I would email support. They'll get back to you.

[regal zinc]

email where?

[violet tundra]

hold on... lemme check...

[stray slate]

I mean the accent thing, he mastered by generating 21987346872346 songs with the same prompt. What he (and I) were looking is to use the same "vocalist" on a new creation, but as I told earlier your solution seem sensible

[violet tundra]

https://help.suno.com/ for basic questions... gimme a second on that email...

[regal zinc]

why do they make it so hard to contact them

thats not a good sign'

[unique stone]

[e-mail]

[violet tundra]

They don't..... their website says [e-mail] for contact, support.

[violet tundra]

☝️

[neat nexus]

Have you read the rules? Rule #5️⃣ explains it clearly...

[regal zinc]

where does it say that?

on the website

[neat nexus]

That is not a Suno.com website.

[jade sapphire]

ok i am curious is it even possible to have same voice / vocalist on every song?

[violet tundra]

hahahahaha

[stray slate]

Apparently not, that was my first question

[violet tundra]

you're right.

[regal zinc]

this is what im talking about..

[neat nexus]

https://tenor.com/view/king-of-the-hill-you-gif-5929698

[regal zinc]

i searched all of suno.com and theres nothing for support

[jade sapphire]

thanks, thought as much since style and lyrics clearly influence the outcome, which in context makes sense to me to have different vocals

[regal zinc]

it shouldn't be that hard to find

[neat nexus]

Suno.com can be contacted at [e-mail] or [e-mail]

[regal zinc]

where are you getting that from?

its not on the website

[unique stone]

On the subscription page:

[neat nexus]

I don't know this sunoai.audio but it is not Suno.com

[stray slate]

Yeah, but the master @violet tundra recomended to just extend to the point of creating 2 different songs, let's see what happens 🤣

[regal zinc]

what a ridiculous place to put it... anyway.. thanks

[violet tundra]

I'm no master kwok. I was just trying to be friendly and offer suggestions that have worked for me since the help desk people seem to have stepped away.

[regal zinc]

it doesn't say support at all.. only billing

[stray slate]

It wasn't ironic, I really liked the proposed solution. Why is everyone on edge today, is friday and our bodies know it 👯

It wasn't, it was a compliment 😦

[unique stone]

Or you can go to Help and type in "contact" and reach this page: https://help.suno.com/en/articles/2551617

[violet tundra]

Ugh. I'm taking a breather. Need more coffee and cocaine to relax me.

Where the hell is The Harbor!? Missing the harbor.

[regal zinc]

what are you talking about? lol

none of what you said has anything to do with the reason I said any of that

[polar spade]

To any interested: As I understand, this is what I know. You would of course verify before relying on my information:

If you've read the TOS, your billing is processed by STRIPE service. You can ask questions about you subscription to [e-mail] or [e-mail] but may need to chat with Stripe regarding recurring billing issues (as per TOS).

Otherwise, if you signed up to the "other" sites , they would not be Suno, if it's not specifically suno.com. You would check their TOS, contact Us area, or look through your bank and credit/debit accounts for who charged your account.

Dont forget, under your Manage Subscription (click your avatar), in suno.com you can scroll down and click that big white button that says Contact Us

As always, read Terms of Service as a first line of defense that you may be signing onto the wrong website before subscribing or buying anything. They may be legit, but they're not Suno.com

This message will likely self destruct within 1 hour.

[empty anvil]

I want to create a relax Flute music but I get Piano when I enter Flute why? is there any other tag for flute

[violet tundra]

some repetition of "flute" might be necessary. Suno seems to struggle with wind instruments. The prompt: [relax Flute music, reeds, wind instrument, flute,] [song about Elves] got this on 4th iteration (two tries) ... which is kind of close... https://suno.com/song/61a387f9-5c2a-4157-9615-d3da9dee92ee

[naive quail]

Are you using basic mode, or custom mode? I haven't had too much trouble getting flutes when adding them to the Style field in custom mode and then reinforcing it with tags in the lyrics. Here's some playful whistling with a flute accompaniment. https://suno.com/song/d6580455-de9e-4a6a-9c59-7706abb9ea4f

[empty anvil]

Custom and in style of music i used Flute

i need only Flute

[naive quail]

If you want ONLY flute for the duration without ANY other instruments, you're probably going to be stuck doing extensions. Suno's gonna fight you.

[empty anvil]

Few days back i got flute music by uploading a small flute music and the extending from there

but still i got some mix in that too

[naive quail]

There's probably not a lot of songs in the training data that are unaccompanied flute pieces, unfortunately. Suno's gonna fight you.

[empty anvil]

https://discordapp.com/channels/1069381916492562582/1083129341870878780/1279119276703354961 I just created this and i extended repeating the music in audacity for 1 hour

[naive quail]

I have one generating now that is an unaccompanied flute. I used "Unaccompanied Flute" in the style, but I think it was from luck more than from that tag because out of 5 gens with that style, this was the only unaccompanied one.

https://suno.com/song/01e32fb6-da25-4253-89be-29b633d4dfc8

[empty anvil]

Suno should add Indian Classical Music too to Data base

[empty anvil]

yes this came very close but music is very simple

[hollow hare]

Great to hear. You can Reuse Prompt that piece and add the complex styles you want (instrumental tags can do quite a lot), albeit it will take quite some credits and time (as the actual chances may be even lower than 1/5)

[hoary otter]

Would have to go looking but I found some great indian pieces in my prior searches.

Even has some interesting luck with generating some of my own, though I got some interesting Indian themed vocal hallucinations

[empty anvil]

Wow thats nice. If you get please share the link

[hoary otter]

https://suno.com/song/23c8c33f-6b21-4e2b-8c26-a6047149a362

mine

With a little luck and the right input I think you can get what you want out of the existing Suno AI

Have some others I like even more but not willing to share those out, may use those in the future lol 😉

@empty anvil what you were looking for?

My style input was both specific and broad in a few tags. I find the specific often get the specific instruments I am looking for while the broad ones at time yield little treasures for something you might not have considered but that may even at times exceed your original intent. But in all honesty. I find I am disappointed far more then amazed. But like the lottery that is what keeps us clicking on the CREATE button and making Suno more revenue so gotta take the good with the bad lol.

[warm acorn]

Hm, is it possible to make a very short complete silent break (for example short before the chorus)? [break] didn't make such a break.

[hoary otter]

am certain it is, have heard them assuming there is a tag

I know break give you space within the composition and usually reduces your instrumental complexity but the sound still remains though lower

"An instrumental ‘break’ can replace a verse standing as its own section, or might be a short bridge in the music. These seem to work best when only one is used at a time, but adding commas inside the prompt may work. Experiment!"

looking to see if they speak to silence

not seeing it in Suno's advertised tags but lord knows their list is maybe 1% of what the AI is capable of

go figure:

**In SUNO music creation, you can use the [silence] tag to create a period of silence in your song. This tag helps to insert a pause or break, adding dramatic effect or simply giving the listener a moment of rest.

compliments of Co-Pilot 🙂

[cinder slate]

hey so i'm poring over the "documentation" (🤣) and the wiki and stuff, and in the wiki it says:

Use AI feedback to refine lyrics and composition before final adjustments:

Discard less successful attempts and allow the AI time to refresh.

Allowing the AI to refresh can lead to better outcomes when adjustments are made.

can someone plz explain what does it mean to allow the AI to "refresh"??

[naive quail]

That's Co-Pilot hallucinating, tho. 😄

[cunning mulch]

So, uh... a friend had something interesting happen in their song... I won't post the song itself, as it's a bit crude, but because of the prompt, it went from heavily-accented English in the song, to this after it 'ended'... I (with permission) Extended off it, and clipped out the spoken part in question

(accidentally deleted wrong line 🫢 )

[naive quail]

If that's from sunoaiwiki.com, I'd just ignore it. That wiki feels like it was generated by AI; lots of hallucinations in there.

[cunning mulch]

oh, SixString, you hear the infernal outtake I had in one of my Extensions?

[cinder slate]

OK so is there any good documentation/wiki/anything(??)

[hollow hare]

This. The unofficial community wiki is https://suno.wiki

[naive quail]

Missed that one 😄

[naive quail]

The one Lumine just linked is a good start

[cunning mulch]

https://suno.com/song/d69c0c43-fa39-4d17-bc31-2f9114f92842

[naive quail]

https://tenor.com/view/mr-bean-mrbean-bean-mr-bean-holiday-mr-bean-holiday-movie-gif-3228235746377647455

[cinder slate]

ok thanks... where do you go after "the start"? is there any community where people are like somewhat "scientifically" trying to figure out the app? cuz it seems like so much of it is rannndom

[naive quail]

lol. wtf suno 😄

[naive quail]

This channel 😄

[cunning mulch]

funny part is it kept the style, even as all the rest fell apart

[naive quail]

that's nuts!

[cunning mulch]

you can hear it if you listen

[warm acorn]

That was the second thing I tried, but it didn't help, it didn't seem to do anything.

[cunning mulch]

it's like you can hear the plastic flesh melting off the metal skeleton

[naive quail]

@charred merlin did some pause/silence testing. Not sure if anything conclusive came out of it, but he'll probably chime in soon.

[cinder slate]

ok cool, that's great, i can work with that... is any of yall working together to do this? like running "controlled experiments" and stuff to build a corpus of knowledge/examples? lots of great knowledge here for sure but i find the randomness of this app really frustrating

[hoary otter]

supposedly [Silence] is a Suno tag

[naive quail]

theharber has credits to burn, so he's done a lot of testing and HAS organized some of it into playlists

I have an experiments playlist, but it probably only has like 1% of my experiments in it

[cinder slate]

ya we've chatted a little about it, i still think that a controlled approach to experimenting is possible and would be useful (or if it's not useful, that in its would be a relevant finding)

[cunning mulch]

so... any idea what may have happened for the language switch in the song?

[naive quail]

Suno.... crackpipe... 😄

[hollow hare]

How about extending that point with just using instrumental tags?

You can also potentially* brute force it by making an extension (set to 3.0), [End] the piece (style: End, Short), combine the song, the continue off*

*Disclaimer: I would be careful of this method, sometimes the song would not want to continue after that point

[cinder slate]

i could put together a sketch of like a "research design" but just starting with basic stuff like a prompt like "music" or "..." and run it 40 times or something

[hoary otter]

well that doesnt surprise me.

I have used END tags and often multiple and still has Suno generate me a 2 min extension that goes into another. It is in those moments I find myself asking, was that intentional as I am forced to click CREATE and pray for a better result. Which maybe we can request a modification of the CREATE tag to CREATE & PRAY

[untold pawn]

Is there any way to recover a song if you deleted it by misstake? 😦 Deleted the wrong one. Still have the link to it open in a tab:/

[hoary otter]

yes

trashcan at the top of your songs list

[hollow hare]

(Via library, see top right)

[hoary otter]

couldnt be more ummmmmm elusive...

[untold pawn]

oooo thank you guys!

[naive quail]

Some of those false ending / variation restarts are interesting, at least. I've uploaded some of those to use as b-sides of the original

[hollow hare]

Yeah, 3.5 really does not want the DJ AI to stop playing. If you want to end the song, you have to set the Extension as 3.0 (or even 2.0) as they understand what [End] means.

Though I like using [End] in 3.5 in some of my generations to hear what other variants Suno makes. Had a few bangers from that which I had extended from

[warm acorn]

Well, I put this on the end, works well for me to end the song:
{end}
[end]
[END]

[naive quail]

hate when that happens. I then have to navigate to ~ page 1200 in my trash and spam the right arrow to go thru the remaining pages to get to the last page. 😦

[cinder slate]

lol first page of the suno.wiki:

There is no external API or mobile app – yet. The website is the only place to get the full Suno experience. Do not enter your login or account credentials in any other website or app.

k

[warm acorn]

yeah, that trash folder is a pain in the ass. XD

I wish there would beside "move to trash" a "destroy" option. 😏

[naive quail]

1320 pages ::sigh::

[untold pawn]

Jesus, im just a rookie here 👀

[warm acorn]

99% of what I move to trash is nothing I would need anymore. And because I sometimes generate dozens of songs without getting something useful it is fast full of songs which you need to delete one by one.

And what I also didn't understand... I created today a complete new song, but in the trash folder are on the first and the last page songs from that and all other from the last weeks in between. XD

[naive quail]

Tho you CAN use the standard windows SHIFT or CTRL selection modifiers for multi-select.

[warm acorn]

Good to know, thanks. But how do you should know? Normally there is a "deleted selected" option or something. XD

[naive quail]

hidden feature 😦

like the resizable lyric's column in desktop browsers. There's no grabber for it, so many people don't actually realize that it can be resized.

[cinder slate]

i feel like everything in this app is a hidden feature... 😢 sorry not trying to be a downer, but an app should have a documented API!! (IMHO)

[cinder slate]

(API in the UI sense, not a REST API necessarily)

[naive quail]

I suspect that documentation would be surprisingly sparse and just leave people even more irritated.

[cinder slate]

I suspect that documentation would be surprisingly sparse
can you elaborate?

like here's the cynical take: "we won't tell anyone how to control the outputs, so to get what they want they will just have to spend loads and loads of tokens"

[naive quail]

The internal list of tags that they've intentionally coded is probably quite small. They're likely as surprised as we are at some of the things that Suno responds to.

[unique stone]

Yeah I have no knowledge of how it's actually trained, but it's known that it's trained on millions of songs and I doubt someone has manually combed through it all and labelled all of the song data with the appropriate tags... so there's probably a lot of other data involved, that the model has learned but nobody actually knows.

One day it'll probably get to a point where it can all be documented and behave in a more predictable manner

[warm acorn]

Uhm... maybe an illusion. 😄

[cinder slate]

I doubt someone has manually combed through it all and labelled all of the song data with the appropriate tags.
i'd be interested to know what kind of training you have in mind, then, cuz often training REQUIRES labeled data

are there any known (self-identified) Suno engineers on this server? (are there other servers for this type of stuff?)

[cinder slate]

like i'm really impressed with Suno, and some of my generations I really love, that's why I'm here... now I'm trying to understand if their business model is like slot-machine-style or what

[unique stone]

It surely does have to be labelled somehow, but with so many songs, like I said, there's probably some other data involved.
A good example is how to get Suno to speak instead of sing: You can tell it you want "spoken word" and "narration", but it will still try to sing unless your lyrics look like a script/book rather than lyrics. When I noticed this I had a 🤔 moment. Like maybe there's a load of text data that gets fed into the training dataset alongside the audio.

I don't think their intention is to make it difficult and slot-machine-like to control. It's just unfortunate that for whatever reason at the moment, it is quite difficult to control accurately. With the state of AI at the moment it feels like the computers are ahead of the humans. We can throw data in and then have to work out how to get what we want back out of it 😄

It will be disappointing if you are expecting Suno to give everything you ask for at the moment. But, if you input stuff, see what it makes, pick what you like, then feed some more input etc. and just cycle back and forth like that, it works well.

[warm acorn]

Ah, that one, I forgot about that.

[rocky zenith]

Suno is the first songwriting AI app I've used and I'm pretty impressed. I've been trying to figure out how to get it to use the type of voice I want--specifically, I would sometimes like to use a female voice and occasionally even a child's voice. The app seems to be male-centric and it's almost impossible (for me, anyway) to get anything else. I got a female voice for a whole song ONCE. I got a female voice for the first verse TWICE and, one of those times, it also used the female voice for one line in a later verse. I tried telling it three times, spaced throughout the description to use a female voice, but that doesn't work. I tried setting off the request with commas or hyphens; no joy. And (separate issue) sometimes, when I ask for "pop" genre, I get country instead. Still--it's pretty amazing, what it can do.

I just tried a new one today. I used my own lyrics and asked it to use an atmospheric female voice. This time it worked. It always makes two songs every time I ask for one and this time it used the female voice for both of them! 🙂 I wonder if it made the difference that I put that instruction on a line of its own?

[fickle lake]

not sure if this is the right place, but here is my suggestion for mp3 download: #💡┃suggestions-web message

[cinder slate]

This is not the case with most AI tools, fwiw, like the chat bots have extensively documented APIs and so on

[cinder slate]

That works well for Suno because I’m burning through credits…

[nimble iron]

Adding to this, I thought that pressing Ctrl+A would also work.
And while it appears to at first, once you click to 3 dot menu on it will unselect them all

[cinder slate]

Even think about midjourney and others, there’s a lot of unpredictability but there’s so many more parameters for you to fiddle with

Suno has like… one button and then cross your fingers

[cinder slate]

I agree it is amazing what it can do, again that’s why I pay for it, but I’m also disappointed in what it cant do, esp if you’ve worked with other gen AI tools

[neat nexus]

What?

[cinder slate]

"Extend"

[neat nexus]

Are you like not using [metatags] at all in your lyrics? Adjusting styles as you go for effect?

[cinder slate]

yeah i'm experimenting with a lot of that, and sometimes it works, and sometimes not

[neat nexus]

Well yeah. It's an Artifical Intelligence. That means it doesn't always do what you ask.

It has a 'mind' of its own sometimes.

[cinder slate]

that's cap, if i use openai API, and i enter the same prompt and the same seed i know what i'm going to get

same with midjourney - Suno does not expose seed data (!)

[neat nexus]

To a certain degree it's like making a soup. You add ingredients hoping that they will blend. Sometimes they do and sometimes you have to add a little more and make adjustments as you go.

[neat nexus]

What I have been repeatedly told by the developers is that the methods that we're using to make music do not correspond to the same methods used in image generation or in video generation. Full stop. It does not work like that.

[cinder slate]

ok, fair enough, is there any public info about these methods, then? for example, Udio does actually give you the seed number

[cinder slate]

like what architecture are they using? it is a diffusion model or what? i'd like to know these things

this is not a "Udio vs. Suno" thing, I think they're both cool, I'm just trying to understand what is going on, AI is not magic

[neat nexus]

My understanding (limited as it is, I'm a moderator, not a team member) is that it is a neural net of some kind. Udio does a transformer model. Which is why they have their own strengths and weaknesses, They approach the problem of music generation differently.

[cinder slate]

transformer architecture leverages neural networks...

[unique stone]

I use Udio as well and whilst they do offer more controls, the output is still - to a large extent - unpredictable.

[neat nexus]

Well I did say my understanding was limited.

[cinder slate]

from what they say publicly, they are "learning the grammar of music", this sounds like taking a bunch of data and building a vector space where you can do semantic instructions and stuff, i will study it more before i weigh in too much, but this isn't something magic, it's big math-crunching like any other AI

[cinder slate]

i agree - a lot of those parameters are equally "automagical" - but at least give me the seed number for g*d's sake 😆

[cinder slate]

no worries

[unique stone]

I've not actually figured out how to use the seed properly in Udio 😅 apparently Suno doesn't have/use seeds (this has been requested before)

[cinder slate]

i think if you enter exactly the same prompt (all parameters/sliders the same) and the same seed number, you will get the same output, but anyway this is not the Udio discord so I'll shut up

[polar spade]

Sunonis pretty good at deciding the most likely intent of your song bases on lyrics also. So it's a combination, or amalgamation of lyrics prompts, metatags, grammar, language and so on that it tried to give itsj best interpretation. Almost as it it say "I know you say "pop" but I can tell in your heart from your lyrics ,that you want country.

[polar spade]

Use. "Kpop" and the first 5 will be female. 🏔

[polar spade]

Wait. That was supposed to be Mt fuji for jpop. Sorry

[cinder slate]

according to chatGPT (sorry lol):

Suno is a music generation platform that uses advanced machine learning models to create realistic and emotionally resonant audio content, including songs, voiceovers, and sound effects. At its core, Suno employs a combination of transformer-based models and other AI techniques to analyze and generate music, lyrics, and vocals based on user inputs.

so i think i need to study BARK and see if i can figure it out

if anyone is interested in doing this kind of technical (as, in, computational) deep-dive, please hit me up!

[violet tundra]

"realistic and emotionally resonant audio content," And the thing is, if you go on Reddit or any musician-based site, you will find the tired "AI will never produce anything with emotion," which is nonsense.

[polar spade]

Chatgpt Doesn't know Suno like we do. It has a heart and soul but can be confused at times

[violet tundra]

As can be most of my extended family. (the confused part). I think we hold pre-sentient AI to an awful standard. A few self-monitoring tokens and better persistent memory, and as some people out there suggest, we would have a type of language-sentient entity.

[cinder slate]

i can't tell if you mean this metaphorically, because some people don't, for me these are computers doing a ton of math, in fact what they're doing usually is pretty "dumb" at the basic level, even though the results are impressive

[cinder slate]

this is an extremely controversial view, but not a crazy one

[violet tundra]

that's a very balanced response. 😃

[cinder slate]

the philosophical underpinnings of some of Suno's responses to the lawsuit are actually really interesting and really really controversial (i.e., they basically say it's "learning" in exactly the same way a child learns to play music) - there's a lot of cognitive science that is being assumed there

[hollow hare]

I would be critical of the answers given by LLMs, they can lie to you with confidence

[cinder slate]

That’s because of tokenization, LLMs don’t process “words” per se

Did you try putting the word in quotes?

I’m doing “regenerate” and it’s sticking with 2… tokens are not equal to words

[safe hazel]

😰

[runic kayak]

Sooo... I tried to use my Mastercard to subscribe for a yearly subscription, but somehow the card can't be validated. IS it because it's a debit and not a credit?

anyone else who knows how to get that problem away?

[gilded yew]

I created songs using the Suno app on my iPhone. I clicked Download but nothing happened. I checked the Download folder in the On My iPhone section in the Files app but it’s not there.

Is there a trick to downloading songs from the iPhone app?

[safe hazel]

Maybe the [e-mail] email address should be pinned

[safe hazel]

You'll probably have to contact Suno at [e-mail]

[runic kayak]

I'll do that

[fallow tide]

Is anyone experiencing issues with audio fidelity dropouts in the middle of a generation in the last 5-10 minutes? I really liked what I got from the output, but then all of a sudden, really poor audio quality in particular areas of a song.

[violet tundra]

So can I.

[astral flame]

Are you proficient with ripx make tutorials about how to use it with suno? I am trying to learn how to get the vocals out of the other instruments. I usually end up removing necessary notes. Do you have any specific tips?

Is it possible to make an instrumental with suno and then have it make lyrics over the same song without it making a new version of the music?

[cinder slate]

great question, would love to know as well. Regarding stem splitting, vocalremover.org does it pretty well, i think kits.ai has a free one too, you could try splitting out the vocal and then feeding the output back in and maybe it works better - i haven't found any stem splitter that is 100% clean, you can use Izoptope RX and stuff but that's expensive and maybe not what you want

(you might think that an AI that is generating the songs would be able to perfectly isolate the stems, but that's not my experience)

does anyone have experience with or use prompt generators? (like not the Suno native one) I'm curious if they're any good

[leaden lark]

From my experience, Bandlab is pretty decent, but it also helps to have a somewhat clear Suno song first before you split the audio as then you get cleaner stems

[violet tundra]

I used a prompt generator a while back, which is now offline, apparently. So, in case anyone finds this useful, I'm posting the template that has really helped me make some distinct songs.

Goes in the lyrics box

<SONG_DETAILS>
[GENRES: Progressive Rock, Alternative Rock]
[SOUNDS LIKE: 80s pop with blend of 90s grunge]
[STYLE: Intense, Atmospheric, Percussive, Gritty, Anthemic]
[MOOD: Dark, Reflective, Empowering]
[VOCALS: Male, Dynamic, Emotional]
[ARRANGEMENT: Medium tempo, Complex rhythms, Anthemic choruses]
[INSTRUMENTATION: Electric guitars, Synth pads, Bass guitar, Drums, Percussion]
[TEMPO: Medium, 80-110 BPM]
[PRODUCTION: Polished, Layered, Dynamic]
[STRUCTURE: Intro, Verse, Pre-Chorus, Chorus, Verse, Pre-Chorus, Chorus, Bridge, Outro]
[DYNAMICS: Tension build-ups, Powerful drops, Emotional peaks]
[EMOTIONS: Introspection, Defiance, Empowerment]
</SONG_DETAILS>

Lyrics go below the above instructions. usually plenty of room for a long song.

=========

Goes in the Style Box

Alternative Rock, Anthemic Rock, Pop, Synth, SEE <SONG_DETAILS> IN THE LYRICS FIELD FOR DETAILED INFORMATION

[quasi summit]

@ the Suno dev team - Please put the navigation back on the Create tab on the website. I made songs that I cant access because you removed the pages and back navigation.

To recreate make 100 songs. Now try to get back to song #1. You cant unless you liked it.

I moved this post to a more appropriate spot. #🪲┃bugs-web message

Oh, google Chrome, Windows 11 Pro

[safe hazel]

You can navigate through all song pages from the library

[quasi summit]

Thank you. Ya I discovered that navigation is now only in the Library tab after posting this. I appreciate the screenshot.

#🪲┃bugs-web message

[naive quail]

This is really just a big tag dump in the lyrics box. Suno may use the tokens that it recognizes, but the rest (likely the bulk of the text) will just be prompt noise. You can completely shuffle around the text inside the <></> sections and still get similar results.

[safe hazel]

does the <> and </> do anything?

[quasi summit]

Good question. Lemme try.

[safe hazel]

you dont have to waste your credits on that if you dont want, I'm sure SixString knows

[quasi summit]

It does navigate buttons and thumbnails. But it doesn't page scroll.

[naive quail]

likely just gets ignored. It's probably no different than using [] or () or ##

[quasi summit]

Oh you meant in song creation. I misunderstood.

[naive quail]

if the text inside breaks down into tokens that Suno recognizes, it may use 'em like it would any other tag

[safe hazel]

do we know of any punctuation that has a specific role like (parenthesis) are reliably back up sounding. I don't know of any others than brackets

[quasi summit]

They're not reliable. per se, but they do work. So does capitalization and ... after a word to stretch it out.

[naive quail]

I've successfully used other random punctuation in the place of brackets before. 😄

like in this track https://suno.com/song/06ff37af-a5b6-4b4d-89f6-d4ece9679c61

[quasi summit]

I used "word" a couple days ago and it flawlessly and clearly said what I had in quotes.

[safe hazel]

I also like cheese btw

[quasi summit]

"LETS CREATE" [female voice] [whisper] for example.

[naive quail]

https://suno.com/song/b6e6d14b-1d01-4556-9ed2-4e54b89c75c1

[quasi summit]

Whisper will be offset by the cap letters

[safe hazel]

Sometimes Exclamation marks give me lyrics that are more scream-like

but, not always

[quasi summit]

Very true. Same here. But not reliable : (

[safe hazel]

suno just like my ex, never reliable 😔

[naive quail]

In my testing, it seems to be more related to where those words fall in the song structure.

[safe hazel]

yeah I was just thinking that as well

I think exclamation marks paired with certain structuring can work more often

[quasi summit]

Working with splicing lately. Creating two songs in two different genres, splicing them in editor, and returning them to the AI for a blended mix. Super fun. I cant seem to prompt multi-genre in one song. Suno hates it I guess. Here is an example of a chant and electric guitar splice (after the merge in AI).

[untold briar]

Under used capability

[quasi summit]

It has been for me. 100%. But I'm all over it now. Super cool potential.

[safe hazel]

the few times ive tried feeding suno back it's own output, the audio quality was just not there

[quasi summit]

I exported those two songs into Davinci Resolve for the splice. It by default improves the audio quality when you cook it.

[violet tundra]

yeah, I'm not saying it's perfect, but it allows me to think of things and articulate things in a more precise way, and if I relax on the need for 1:1 input-to-result ratio and focus more on getting the desired effect after a couple hundred credits, it does work--even the bpm and tempo and some of the largely ignored commands. Which is nice, because once I get that one working sample, I can extend and get the unique song I was after.

[safe hazel]

oh that's much fancier than i am. I just started using audible for the first time

[quasi summit]

Blackmagic Davinci Resolve is free. Very powerful art suite. Amazing for art, videos and music.

[naive quail]

I love doing that, tho. I've had some entertaining b-sides that came from me trying to end the original track and getting one of those **false ending / variation restarts ** I clip that bit out and reup it 😄

[quasi summit]

You can also VirtualDJ. It has a free version. But the quality of Resolve is 100x better.

[safe hazel]

oh yeah, sometimes those restarts are better than the first song

[graceful condor]

/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

[naive quail]

Totally for it when used in that context. There was a period a little over a month ago where a TON of tracks were using that and newbs were thinking that Suno was interpreting it all as actual instructions. 😦

[naive quail]

This is one of those https://suno.com/song/fc6e4de7-cd8d-4806-a80f-45ab3e9e544c

beware the jumpscare

[safe hazel]

oh jesus

[naive quail]

born out of a never-ending ending from this one https://suno.com/song/3b8545e8-3f59-40ba-9499-4c0cb8d3a38f

[safe hazel]

i lowered that audio just in time

[naive quail]

hahaha

[whole comet]

Anyone know how you can get an instrumental composition which is just one instrument?

[naive quail]

luck and many credits 😄

[violet tundra]

I can see that. It looks official / as if it should work all the time. Basically, I bounce around. from the basic input with minimal input to the custom input to this when nothing else will work or if I want some unique blend. It's one imprecise tool among many.

[safe hazel]

does [solo] help with that

[whole comet]

it hasn't for me so far - neither has been putting an [Instrumentation] tag in header

[naive quail]

I think 'solo' mostly gets interpreted in the context of 'guitar solo'

[whole comet]

I guess I could take a minimal composition and pull the stems apart

[naive quail]

I did get an unaccompanied flute track earlier by prompting for "unaccompanied flute", but I think it was more luck than suno properly interpreting "unaccompanied" https://suno.com/song/01e32fb6-da25-4253-89be-29b633d4dfc8

[whole comet]

Yep! Was looking for low D pennywhislte peice 😄

[safe hazel]

maybe if you fed it something as an upload that was a single instrument

[naive quail]

Yea, that's my only beef with using stuff like that... it's easy for people to get confused over it.

[whole comet]

Well hell. It's be nice if it WAS official. It's a real struggle trying to get the system to "behave" when it's not even consistent following it's own tag system 😐

[naive quail]

This is one of the <song_details> tracks that I was playing with when that prompting method started popping up https://suno.com/song/879ced57-1f2d-4dcf-8270-6ab2169f7945 I totally scrambled it all and then spread it all throughout the lyrics 😄

haha. I forgot about adding the solo 😄

[safe hazel]

something that ive seen in my songs is that when I have some bracketed prompts in like the middle of the song, suno will still use those in the beginning of the song. Like, it reads the prompts of the whole thing before starting to form the melody it uses and so later prompts are still incorporated throughout the entire song.

Is that a thing people agree upon?

[naive quail]

definitely. I've noticed this, too

[whole comet]

So - best to extend and add the tags only to the relevant extension prompts, to limit specific effects?

[naive quail]

from a conversation that I had with @charred merlin

[safe hazel]

now we just need to figure out how to tell it to keep certain prompts localized to where we want them

[naive quail]

if only 😄

[safe hazel]

but, i do like how it incorporates all the music prompts into a cohesive melody

at least sometimes 🤭

[naive quail]

always keeping us on our toes!

[whole comet]

This isn't a complaint - it is what it is - but it's sounding/feeling more like Suno is - so far best used as a raw source material generator in a production pipeline, than as something you can ( easily, practically ) use for song production. Pump out a double handful of tracks, use other tools to tear them down into components, then mix and compose those parts. But all that is happening downstream outside of Suno. Don't get me wrong - the double handful of raw material is a huge leap forward in the process - so it's still of value 🙂

I look forward to see if/when/how-much-of that pipeline Suno willl eventually be able to take over 🙂

[safe hazel]

I expect someday you'll be able to specify exactly where you want certain instruments to happen and even what notes you want it to play. Someday far into the future

(but maybe not that far)

[whole comet]

Maybe not that far 🙂 Especially looking at the rapid progress of AI tools in other venues.

[naive quail]

Yea, it really wasn't that long ago that MidJourney was giving us mutilated flippers instead of hands and feet

[whole comet]

Heck - ChatGPT now even can create palatable recipes 😄

[graceful condor]

@whole comet The speed of rapid progress is a little...ah..suspect, I sometimes think. One can't help but wonder if these mega-corps were working in the background and signed everyone to strict NDAs with handfuls of money...

[safe hazel]

I first discovered suno only a few months ago, and i was just completely floored by how amazing it was. Now im like "hey do more 😒 "

[whole comet]

There's def an elment of that! "But I want it to be perfeeeect!!!"

[cinder slate]

Again… isn’t it kind of odd that you have a machine that can generate a multi-genre experimental structurally complex composition, but can’t do “please play the sound of a woodblock once every quarter note”. I.e. it feels like you should be able to solve for the “simpler” kind of sounds than complex ones. Just my thoughts.

[whole comet]

If it was programmatic or algorithmic - yes.

[cinder slate]

Or if it “understood” language like LLMs do

[graceful condor]

I have a theory on this. The complex compositions might be easier because the dataset is comprised of finished compositions. You notice how many genres these days are quite formulaic in their bridge, pre-chorus, etc. Suno will even "place" the lyrics correctly based on the tempo, etc. However, a single sound is a data with no "reference point" and so finding source or reference data for that is harder.

[cinder slate]

Agreed. Still kind of puzzling for me

[whole comet]

The very fact that you can meet an 8-year-old violin prodigy who can't tie their shoes, shows that learning systems don't learn evenly 😄

[astral flame]

I use ripx which is why my question is about if there is a guide to using it better. They have all been the same outside of them. I like them bc they separate instruments out further than most others .

Just looking for a ripx guide to cleaning up stems

[cinder slate]

Yeah idk sorry I just heard about ripx, look forward to playing with it

[fickle lake]

#💡┃suggestions-web message
did anyone see my msg?

[whole comet]

I actually have had much better structure behavior over the last day or two, by being "serious" about composition, and ending up giving Suno more context cues. E.g. - have consistent metrical patterns in parts of songs ( verses, chorus, bridges ), explicitly laying out chord progressions, key changes over different choruses, etc.

Given structural cues like those, Suno has done a lot less structural scrambling on me.

[cinder slate]

Not asking in a critical way just curious - have you tried things like many many generations of the same prompt with this structure? Or like using the same prompt but without the context clues? Just trying to understand the reproducibility of this

[cinder slate]

(And whether I should burn my credits experimenting)

[tranquil dune]

how do i make sure piano comes in first

[safe hazel]

you can't ever make sure of anything, all you can do is try. Something like [Piano Instrumental] or [Piano Interlude] as the first prompt in the lyrics can help. Putting "piano" in the style box can help. But, nothing will make sure it happens

[tranquil dune]

Oh okay thank you so much

[tranquil dune]

https://suno.com/song/1f6afd87-7d82-4daf-9e16-320cf99c978d

[safe hazel]

Seems to have worked pretty well. It came out great. I'm not sure if putting "Female" in the lyrics box is the best way to try to ensure female vocals or if putting it in the style box is. I always put "female/male vocals" in the style box so i haven't had much practice using it in the lyrics box, and from what i've seen here people seem to get the best results from putting it in the style box. Just an fyi (but others would know more than I)

[cinder slate]

So I guess my question is - is that because Suno can’t do it or because it’s not documented…?

Like IMO I’m confused why the use of meta tags is not in any official documentation. Unless it’s also part of the generative stuff and there is no “list” of tags.

[safe hazel]

Well, there's definitely a dearth of documentation. I'm sure there are prompts and character usage that can get more reliable results than what we're doing. But, I guess unless 100% of the material it learns from uses these things in the same way, the AI will always have different routes it can go, since it's learned that (parethesis) can mean back up vocals, but other sources may have them doing other things

There is no official documentation

[cinder slate]

That’s exactly my gripe lol

[tranquil dune]

I understand that completely I've tried making a good duet didn't come out well

[safe hazel]

yeah, duets require you to pray to whatever god you believe in for a miracle

[tranquil dune]

FRRRR

[cinder slate]

For context, I’m an engineer, so suno is an app, and it doesn’t have documentation - that’s… um interesting lol. But anyway that’s kinda where a lot of my questions are coming from, I can relate to what you’re saying as a user

[safe hazel]

it's kinda weird there is, well, no documentation at all. But, it's also still pretty new. But it would be nice if they could just give us a teeny little bite of information to work with. But, everything has been user generated. As far as I've seen, devs dont really talk to us

[cinder slate]

And yeah so what do you make of that? Like if I buy a VST for example it comes with a manual that tells you what all the gadgets do. Or many other examples. I guess I’m trying to decipher why they don’t have documentation.

If I wanted to be provocative (just for fun) - either (I) there is internal documentation, in the sense that the engineers know roughly how suno behaves and what it responds to vs ignores, (ii) there is documentation but also extremely high randomness and “temperature” in the model, or (iii) there is no documentation, and they know about as much about its behavior as we do (I mean prob loads more but in the sense of just “observing” it) - apologies to any devs reading this, I have no disrespect for the technical implementation

Sorry I’m not trying to rant in this channel or anything, I have a serious desire to know the technical details that’s all I’ll be chill 😂

[safe hazel]

i doubt devs read this

think of it as a prototpe though. It's unfinished and so why make documentation that you'll eventually have to change. We're getting in on like a beta edition of something (or alpha maybe) so it's all wild west until the model is more mature

but im sure there are lots of reasons why they haven't said more about how it works

[cinder slate]

Yeah that’s fair. I was kinda using “documentation” as a shorthand for like a basic knowledge of the rules of the game (how does it respond to tags, which tags does it recognize, etc.). It’s unclear to me whether the internal team knows this and I’m curious about that

[safe hazel]

they have to know something about how it works, since they know how it learns (which, we dont).

[cinder slate]

Indeed but what about how it will behave? Do they have systematic knowledge or is it based on observation? Or both prob

[untold briar]

The only way to be 100% sure is to generate the piano first, by itself. Then Extend that clip with the rest of your song.

[untold briar]

I don't know if there is an official list or not, but my guess is there isn't a super useful list, at least in Suno up to this point.

For example, you might assume tags on this official page https://suno.com/explore would be 'official' in some sense. But it's not difficult to find tags on that page that don't work at all. And on the other hand, it's not difficult to find tags that work really well, that are so weird and random they couldn't possibly have been created on purpose. Browse some @alpine venture songs https://suno.com/@Yolkhead for example.

The general feeling I get, pure vibes, is that Suno was initially not thinking of very complicated style prompts. Especially about complicated [bracket stuff] in the lyrics. But the users tried stuff, the model did surprising things, and it's only somewhat recently this tye of capability is possibly being leaned into more deliberately. But who knows.

[untold briar]

This was one of the first things I tried in Upload. I need to get back to it. There is a lot of with the clip lengths and how you structure the spliced songs. I had the best luck using some separate pieces of each clip in the beginning of the Upload, and then literally overlaying the songs on the last part of the Upload. But finding the right lengths for the chunks and the overlayed section is tricky.

[cinder slate]

that sounds like a normal and healthy process of prompt engineering, chatGPT went through the same growing pains

[shut moat]

if it was everything rolled into one they'd be charging a whole lot more money than they are for a subscription.

think if they kept the price the same 8 / 12 per month for a fully functional music production studio.

they'd put the rest of thje idustry out of business. I think AI is an excellent brainstorming tool. Maybe real musicians want to use it to explore ideas, then hire the musicians they'd need to execute their idea.

and then there's the hobbyists. who don't have the skill to make the music they have in their heads.

I make a living off of art, like physical mediums, but I use AI to brainstorm ideas or get a shape started for abstract work.

[acoustic hornet]

Has anyone had any luck making something that legitimately sounds like 80s pop/rock - something in the vein of Cyndi Lauper or The Bangles?? If so, what kind of prompts did you slap in the style section?

[shut moat]

it's funny, that one of my clients actually introduced me to Midjourney, she has aphantasia and we were able to work together on a comission she wanted since she couldn't "see" what I was describing to her.

I put feelings and emotions into my style section maybe it helps, maybe it's placebo.

I also try to direct the song in the [brackets] with the lyrics.

actually helps with the song's structure. I'm not sure what the general flow of an 80's pop/rock song is since my cone of interest is metal/punk/skacore/hardcore

[acoustic hornet]

I modeled my lyrical structure around a Cyndi Lauper song, specifically the song "Good Enough" from The Goonies film. The prompting of "80s pop, 80s punk" would seem proper, though I am getting stuff that continually sounds like modern day boy-band.

[safe hazel]

well what's the difference between 80s pop and today's pop?

[acoustic hornet]

I see what you're getting at with that query, and I've attempted including specific instruments and "moods" in the prompt, such as "synthesizer, upbeat, high energy" - still get boy-band-ish sounding music.

I even had one that decided to pick up only "80s" and "rock" from the prompt, and I got something that sounded like Motley Crue.

[safe hazel]

there was someone in here before looking for prompts for 80s rock and they said ultimately these style prompts worked: 80s Glam metal, 80s Hard rock, classic 80s rock (i dont know though, i havent tried them)

[acoustic hornet]

I have used the "80s glam rock" in a couple as an attempt, and while closer, for sure, I still get these really "phat beats" that are common to todays pop.

Not that this is the place for suggestions, I do kinda wish there was a "negative prompt" option within the interface.

[safe hazel]

Paimon says they use this site for style prompts, here is cyndi's page https://rateyourmusic.com/artist/cyndi-lauper

maybe try 80s synthpop

[acoustic hornet]

Ah, Paimon. I've spoken to them before. Good call. I'll definitely tickle that option.

[shut moat]

Have you tried locations as a style modifier?

I made a sort of chanting, war march song, but to get the choral section i put in "black cathredral hell march"

https://suno.com/song/8e827c52-9e37-4011-b459-0f667411ea6e

[acoustic hornet]

@safe hazel the inclusion of "synthpop" really made a difference. I feel a little bit like I must have forgotten my helmet on the short bus now for not thinking of that. Thanks a ton.

[shut moat]

and I rammed "grim dark" down Suno's throat.

[safe hazel]

no problem, and ive never heard of "synthpop" before i started using suno so dont feel bad

[shut moat]

I think sometimes you need to say the same thing a bunch of times.

similar to some AI text to image gens, you need to reinforce the idea with the same words againa nd again.

to really just hammer in the idea.

[acoustic hornet]

@shut moat I agree with that; the multiple inclusion of the same thing, just to get the AI to understand. When making a "viking metal" ballad a few months back, my final prompt included "viking" "metal" "ballad" and even "medieval" each at least twice before I got a sound I wanted.

[shut moat]

I also have the lyrics written in like Latin, or Norwegian, or I convert it to an simplified language than back into english.

the lyrics generator in english is pretty terrible.

it wants to rhyme everything. but does NOT do complex rhyming schemes.

[acoustic hornet]

I purposely write all my own lyrics, often nitpicking syllable counts and flow long before I ever even load up Suno.

[shut moat]

like that chanting song, I used a translator to Khazdul, the Dwarven language in Tolkien's universe

I write way too much so I translate in google to a simple language then back into the suno lyric engine but convert it to a germainic langue then back into enlgish.

lol

[acoustic hornet]

I recently did a tune that included musical terminology like "duolo, cantando, dolore" and more, and to get the AI to sing them proper, I had to write them out phoenetically.

[shut moat]

english sucks. to be honest.

suno knows A LOT of languages.

it can't translate into dwarven but it can pronounce it, I'm trying Klingon next.

[acoustic hornet]

It does, but when most of the lyrics are english, it tends to stick to english pronunciation, at least in my creations.

[shut moat]

I write a lot of stuff in indonesian, it pronounces that almost perfectly

[acoustic hornet]

Even the simplest of terms get goofed on occasion, like writing out the word "cause" as the shortened version of "because" that WOULD sound like "cuz" - Suno will still often pronounce it "kawz" - so I usually just write cuz, then edit the displayed lyrics later.

[shut moat]

gauze too might be close enough

[acoustic hornet]

Not talking about rhyming it, that's easy enough, just having the word in there is often mispronounced.

[shut moat]

i think suno is flawed, like it you stuff too many syllables into the bar it will compress the words.

[acoustic hornet]

Just this evening, I'm working on one that has the word "wicked" in the lyrics, and one of my creation attempts actually pronounced that word as "wikt" - like I was talking about making a candle. LoL

[shut moat]

I don't know enough abvout music to know what the maximum limit is before it starts messing with the pronunciation of the words.

could be just a glitch.

[alpine venture]

In general, I think the best place to start when learning a new prompting system is with prompts that are well established, even if they're insane, and then adjust various pieces of it to fit your desired style. Jonathan has lots of good ones, as do many other big Sunoers and some less well known.
https://suno.com/playlist/faeb1142-531c-47b2-b26e-1f0a99b846de

This has some of my personal favorites.

So I'd start by taking the styles here, and tweaking them. This is how I learned sdxl prompting anywho, and it worked out p well for me.

[alpine venture]

Above is on response to this

But yea, the prompting for 3.5 has really taken off lately as people got off their high horses about cargo culting or whatever it's called

[shut moat]

what the hell is cargo culting?

[alpine venture]

Originally refers to just post-modrrn sort of spiritualism, but basically its a sort of denigration of promoters almost "mystical" prompt techniques, typically under the assumption that it is mystical, and not directed

In my case, I'm literally just applying a well known technique used in sdxl, which despite the success of fooocus still gets called cargo culting lol

Basically, in sdxl you want to launch a bunch of neutral-positive tokens at it with you output to smooth out overfitting

With Suno, this doesn't work because the positive-negative style tags aren't 100% accurate

Which makes sense, as "beautiful" in music is harder to establish, given some genres are intentionally hideous

I mean, I guess it's the same for visual, but I use the same technique, which is mixing in negative+positive tags to avoid certain biases, so things like "awful f---img music, broken composition, Mozart"

Will yield, in my experience, better sound than just "Mozart"

Bc it generalizes better

Although, Mozart is not the best example, as he's likely highly overfit, you'll need a lot more than that

[charred merlin]

I love how your brain works, dude.

[whole comet]

I've done many generations of the same lyrics with changing chord patterns, and changed the chord patterns to correspond to different key modulations. I haven't tried this with a broad variety of lyrics or styles. It's just possible that I merely lucked out with the lyrics.

Further experimentation is warrented 🙂

[whole comet]

The term comes from the post WW2 South Pacific "cargo cults" - where native islanders noticed that when the US troops came to their islands, and built airfields, then cargo planes full of all sorts of wonderous goods landed on their island.

So - they build airfields, out of thatch and bamboo, in order to "attract" the cargo planes and their goods back.

Despite being a real phenomenon, it's become a metaphor for trying to manipulate a system based on observed correlation, when you don't really understand the causality of the inputs and outputs of the system.

You know - like trying to determine how to make a completely undocumented AI song generator do what you want 😄

[astral flame]

can i make an instrumental in suno and then make a version of the exact same song with lyrics?

[frozen pivot]

Suno is a playground 😊
Let us know how much you are playing

[frozen pivot]

Not exactly!
But you can upload 6 and up to 60s of what you created for Suno to try to replicate the sound.

Just upload and extend…
Make sure it's version 3.5

[tranquil dune]

Do you think this is good https://suno.com/song/08a9c816-5333-45cd-8b37-55b2d429ce73

[frozen pivot]

What are you looking for?
Sound great…

I personally see no issues with the song...

There is a separate group for showcasing songs 😊

I entered a song during the summer of Suno Vol 2. I won a prize to pay for a whole year of the Pro Subscription 🙏🏾

It's a lot more than the song…
Here is the bare-bone song without any modification

https://suno.com/song/d17a2cfd-7096-4ced-a1fc-0a54febc231d

Let's innovate and Sunovate!
It's our time…

Note:
Custom lyrics
Also added a live guitar to the final production…

Suno is a PLAYGROUND

It's your turn to enter into the competition 😁🙏🏾

[tawny mantle]

Please give me some advice on how to make sounds and voices sound clearer, without interference.

[sterile yacht]

Put a music style known to have clear vocals first in the styles section can help ( even just putting speech or ASMR) works for me sometimes but no guarantees as it all depends on the models mood (aka luck)

[south willow]

hi guys, is there any possibility to insert negative prompting in Suno?

[merry nymph]

Does suno understands in a prompt BMP and B♭ major/B♭ minor?

[runic kayak]

Small update: Today it worked, was able to get yearly subscription. Was just about to send the email and wanted to grab a screenshot of the error. This time the subscription didn't make a fuss and went through without any problems.

So if anyone has that same problem with their card, just wait a day, try again and it might work.

[polar spade]

There is that point that today's pop obviously evolved from 80s pop. But at minimum it 80s was whatever sound could be made from the emerging synths, from Yamaha , Korg, Moog. Lots of organic sampling and experimentation. 16th beat stacatto sound certain was prevalent. Sinwave and digital sounds emerged as well from the mid-70s and progressed as PCs became common after 1983. But it's trending now for today's pop to retro those sounds, sooo... yea.. sometimes little difference but the production and melodies perhaps

Hypogagic is supposed to be , as I understand, the term for emulating some of that digital beeps and boops from the 80s.

[violet tundra]

I wonder if [-.5] would work - similar to stable diffusion strength prompts. Not sure I would waste any of my credits experimenting, but it's interesting to wonder what the LLMM (large language music model?) understands.

[violet tundra]

No solution - just a workaround... Extend off a clean iteration. Pull the clean result into a multi-layer sound editor. Extend again off that same clean iteration. Again, pull that result into the editor. keep doing this until the song is done THEN tie it all together with a clean instrumental backing track with the same beat. That's what I do when I absolutely want the end to be as clean as the beginning.

[charred merlin]

@south willow Not implemented, and from everything we've noticed so far, Suno doesn't respect any form of negative phrasing "no drums" or "less drums" is the same as saying "drums please, and whatever style the word less nudges you toward, thank you 🙏"

https://discord.com/channels/1069381916492562582/1253341197917884530 currently a known-suggestion that the devs are aware of, though.

[cold zenith]

i hope the voice will be more clear in V4. because for some language, it's difficult to understand.

[small obsidian]

Is there any way to get single words from Suno or like conversation without music?

[hollow hare]

From the top of my head, you can use these in the style prompt or metatag to achieve the effect (style prompt to affect the entire piece)

• Narrator / Narration
• Spoken / Spoken Word
• Standup Comedy

A few others that sixstringtoker uses as mentioned by theharber at #📚┃suno-school message

• audiobook
• instructional

[rich timber]

I'm back!

How do you fix a video where it's skipping and the singing part drags?

2. Is it true, that you can upload your music to YouTube, Spotify, etc as long you have a Premium Suno account and you tell your audience about Suno?

[whole comet]

https://help.suno.com/en/articles/2410177
https://help.suno.com/en/categories/550145-rights-ownership

[rich timber]

3. Does anyone mind collab with me on my new songwriting work?

[hollow hare]

Honestly I don't think you can AFAIK. Typically, from what I noticed, the scrolling tends to stop when it comes to a hallucinated repeated verse. From what I heard, regenerate no longer scrolls too.

The best you likely can do is CC the video or make a Music Video out of it

[hollow hare]

You are much better off asking about that at #💬┃general-chat

[neat nexus]

Remember to tag your YouTube videos with AI generated content. It's the rules.

YouTube's rules, just in case that wasn't clear.

[rich timber]

I went to the Gen. Chat. It wasn't useful. I need to collaboration to learn more.

[fluid vault]

❤️

[neat nexus]

Just to make it clear.

YouTube in its wisdom decided to hide this stuff behind multiple links. leading to confusion on the matter.

[glacial scarab]

yep super "smart" decision

[neat nexus]

Right?

I don't know why they did it that way.

[naive quail]

Single words? Like "banana"!? https://suno.com/song/07fc39bb-0312-4db7-978a-32d1a828cb5c

[neat nexus]

Most people miss the first link when they're uploading so they never see that they're supposed to click the message. Then those who do read the surface level and never click more. Those who click more, usually read the surface of that and are reassured, because it doesn't mention music or because their video is clearly unrealistic so they think they're fine.

Only those who dial down into it see that there's an issue and yes, we do have to declare it. It's almost like a hidden IQ test or something to test how much people do their homework on this.

[polar spade]

Metaprompt as

[Spoken] Hello, how are you

Seems to work for me without a line-down before texting the line.

If you then add a linespace, Suno will often then sing the following lines until you prompt again.

[glacial scarab]

Yea that was exactly how i went through it. "Ah man i have to check this.. oh no i am fine.. wait there's more.. oh no i am fine... wait there is more... Ah man I have to check this"

[neat nexus]

Exactly my experience too.

The worst part was telling others about it because they were sure they were fine.

No, dude, you need to click more. And then look at this.

[glacial scarab]

yep and then.. (I don't if it is by now) it was not possible as a mass action when was rolled out so i had to edit every single video this applied to manually..

[neat nexus]

Argh! Yes! So annoying!

I mean I did it but still...

[near flume]

Vocal clarity depends a lot on music and voice intensity. You can have a song start of clear but end up getting distorted the further it progresses. This is a known bug and Suno is working on this,

[oak meadow]

So I just bought a subscription and I was under the assumption that when you pay for the mid-tier subscription ($10 a month) when you would generate/create projects It would generate 10 of them? Is that not the case?? I'm just trying to make sure I'm not spending coins and only getting two tracks per generation

[oak meadow]

I didn't know where else to post this question there wasn't like a help channel

[near flume]

It generates two versions per click (10 credits). You are purchasing 2500 credits which will give you 250 "clicks" per month.

[oak meadow]

Okay thanks for clarifying it just seems like the subscriptions offer more generation because of how it's listed on the website.