#๐โsuno-school
1 messages ยท Page 51 of 1
no link to cancel its a good thing i mailed billing since its a second free account that was created by accident
Ahhhh, deleting an account, not cancelling a subscription.
I remember someone was asking about that a while ago because of the GDPR's right to be forgotten, and as far as I know the only way to go about it was to specifically request the account being deleted -- but I'm going to sprinkle a heck of a lot of salt on that, because it's been a bit since I've looked into the matter โค๏ธ
yes thats what i meant sorry for the confusion
if you want to quickly use your phone number on your other account in the meantime, you can go to your Manage Account settings and connect the account to a different provider instead which should then allow you to remove your phone number as the primary login method to use it on your other account instead (but still email to get them to delete the accidental one once they get round to it)
my original request to suno was to merge accounts since they are both free and linked to my persona but that request is a lill too tricky for suno to handle since it could be a malliciouse attempt to take over anothers account
i doubt that works like that
i have just tested it and it does
since the new account is already linked to my phone number i cannot use my phone number on any other suno account
What do people think of the new Ideas section on the Create page? It's stopped me from being able to use Suno
the new suno account was an accident so i need it deleted
thats why you need to remove the phone number from the new account so you can then use it on your other account like i just said
like i said there is no way to remove the phone number since its the only sign in method for the new account
that i need removed
^ add another provider in the meantime
i'd rather not that would look like im using two free accounts harri
thats fine, just trying to think outside the box 
no thanks i like my box
#๐งโsuggestions-layout if you want to talk about why you don't like it, or #๐ชฒโbugs-web if it's actually preventing you from using the site.
I wanna get male voice I wrote many thnigs but still I get girl voice what should I write?
I have posted in both and sent support@ an email
@jagged narwhal what types of things have you tried, so we don't suggest the same things to try ๐ -- there's no 100% guaranteed method, but you can try to figure out what prompts could be helpful based on what you're looking forward and what hasn't worked so far โค๏ธ
Can you backdoor your way to the custom editor by selecting Extend on a track in your library and then spamming tab 21(ish) times to get to the "Clear" button to reset it from Extend mode back to Create mode?
I can't get Voiceover to recognise elements on the page since they added these Idea thingies, which makes it impossible to do anything useful on the create page, it was working fine before they added this extra unnecessary feature
I looked at the suggestion-layout, is it an accessibility / ease of access matter, or are you unable to open the "custom mode" section?
Suno seems not to care about people who are disabled
I can't do anything on the page without getting stuck in some form that Voiceover can't read to me because the fields aren't labelled
I actually upgraded from Monthly Premier to Annual Premier last night and now I just can't use it
if you're able to get to any extend button in the library, that will shoot you right over to the custom editor, but you'd still have to Clear the extension to reset it back to the normal Create mode. Looks like it's about 21 tab presses to get to the clear button once on the page.
that sounds really impractical, how am I supposed to count 20+ tab presses on a page which gives me no feedback about where I am?
It was more of a hack to get around the current issues while they take another two steps backwards. I wasn't sure if it was just the simple mode page that was the issue or if the custom create page has the same issue. Visually, custom mode doesn't have those ideas sections... but if it still has no labels, that's obviously a no go.
I should have twigged how careless Suno is when they did the UI change which put me out of action for 2 weeks while I sorted through issues. Stupid me subscribed to Annual Premier a day before Suno goes and wrecks the whole user experience
I have 8060 credits that expire in 5 days because I haven't been able to use the platform properly
@vestal rapids I completely understand how frustrating that must be -- I know that it must be particularly frustrating if the UI changes are actively making the site harder to use for visually-impaired users.
Reading your post in the suggestions-layout page helped shine light on the issue -- I apologize if the initial response seemed dismissive, it was a knee-jerk reaction from the previous UI change when people threatened to stop using the website because the image ratio had changed. ๐
#๐งโsuggestions-layout message I came across this comment from a couple weeks back where @prime gale had acknowledged that the team needed to make improvements regarding accessibility. I know they're working to get things right, but I'm sorry that you feel like things are taking steps backward before they move forward. โค๏ธ
(apologies for the ping Franklin, I just wanted to help amplify @vestal rapids's pleas ๐)
Suno can be slow to respond, but from I hear from other users they do make things right in the end. Obviously it's not just about the credits, you want to make Suno songs now and you can't. And you want Suno to avoid doing it again in the future. I can tell you multiple mods/guides have highlighted your specific problem to Suno folks, it's on the radar for sure.
I love Suno, I desperately want to use it, it was working for me yesterday (UI issues aside) but this most recent thing has been a breaking change, it's just not working
they should roll back the Ideas section until they can come up with a solution
Unfortunately I think many visually impaired people will just give up without making noise, but I am the kind of person who makes noise because having disabilities shouldn't limit my access to the world
I don't want to speak out of turn, as the support team will surely have a better response, but if you're not satisfied with the service, and especially if you're unable to use the service, I'd look into adjusting the subscription plan until things are resolved -- that way you're being proactive, and can try to avoid any complications later if you feel like the credits should have been compensated after the fact. ๐ฌ
I am not worried about the money or the credits
I just brought up the issue of my 8000 credits to highlight how little access I have had
the month before I burned through 10000 credits and bought extras ๐
honestly I don't know how to do that, it's something I never have to do on any other sites
let me try
No I got stuck at "Let's make a song to cheer your country on at the 2024 Olympics"
and some form I can't read
I can't even escape out of it
previously I had a text box for the Suno prompt
now it's just broken
the frustrating thing is that I use complex apps like Logic Pro without issues, here I can't even use a website
From the homepage, press tab 3 times to get to create, press enter
press tab 10 times to get to the custom tab, press enter (7 times if you're already on suno.com and click the nav bar on the left side for context)
press tab 5 times to get to the lyrics box
press tab 3 times to get to the style box
press tab 10 times to get to the title
press tab once to get to the create button, press enter
By NO means a preferred solution, but hopefully it can help you get back to creating songs in the meantime ๐ฆ โค๏ธ
want to see annoying?
It'll refresh back to 50 in 11h30m
Oof, you got that little 5 too ๐ค
It's 02/08 for me 
I gotcha -- feels useless since it costs 10 credits to "Create", but it's like how a $1 bill isn't useless if you want to buy something that costs $2 -- you're half the way there if you get your hands on more currency in the meantime โค๏ธ
ok I tried to follow this but I can't help being dumped into an Ideas suggestion form
You did this from the Suno website, right?
yes
How do typically navigate with Apple Voiceover? I'll be back home in an about an hour and will be able to check myself on a mac laptop. I might be able to find a tiny tweak you can use in Safari as a temporarily workaround for you and any other users. Either a Safari extension that allows custom website styles, or a bookmarklet.
Voiceover tells me where I am, I can see some large UI items as I am vision impared not blind, the problem is the tabs keeping track of them to avoid being dumped into one of the numerous suggestions
If they had a toggle to disable Ideas on that page the problem would be solved
but at the moment they dominate the UI
I know there's numerous ideas forums on the discord, but I'm not sure how that would be accessible from the website off the top of my head. This might be something for Suno Support to properly advise on ๐ I hope you hear back from them soon โค๏ธ
once in the form you can't escape out, the esc key does nothing once you hit it
so you can? write instrumentals by just describing the parts of the songs in the lyric section right?*
im still new at this, and haven't figured out how to write beats yet.
it still doesn't know weird / archaic instruments yet. I can't wait to make some hurdy gurdy shit.
Yup! My "serious" songs are all instrumentals, and only use the style prompts and [metatags] -- they work like localized style prompts
Sometimes it can add an overall feel, sometimes it effects just the upcoming section -- Suno kinda does what it wants ๐ โค๏ธ

https://suno.com/song/1e7ca6f6-50ae-43e9-b888-d994c1ce0d6f
It doesn't yet fully listen to me, or im yet not smart enough to make this stuff accurately.
burlesque french math folk, orchestral classical metal, slam polka song. Listen and make your own with Suno.
Support replied, they acknowledged the problem, and have addressed this to my satisfaction, I won't paste the response but they know there is an issue and want to work to address these issues, they just don't know when.
does anyone know if they will have weighted terms?
to at least get acknowledgement is a huge deal for me, I feel seen which I didn't before the reply
tokens, like the I can weight my styles so I can give importance no matter where it is in the prompt.
I do need help, any advice is appreciated greatly
I know it's something they take seriously, as they're all people too -- the only reason they're doing this is because they all love music as well, so it's likely not a matter that they don't care, but rather there's a lot of gears turning at a lot of different rates, so I'm sure that it's just a matter of trying to handle things as best as they can.
Not to discount how frrustrating it must be for you -- but seriously, you're 100% right on the nose that 9/10 people will just take the hit and be upset about it, without taking a collaborative approach in trying to solve the issue.
You rock for your continued efforts, not just for yourself, but for others who might have similar issues accessing the site. ๐ Thanks for trying out some of the solutions that we threw against the wall in the meantime โค๏ธ
The closest thing to weight for now is simply repeating tags more times to make them stronger. To make something weaker is tricker. You can either repeat the other tags to make them stronger. Or (less reliably) you can try removing the too strong style tag from the style prompt and only using it in [brackets] in the lyrics, typically on top.
Actually there is an edge case. For styles that are compound words like "Acid House" then "Acid-House" often works better. Maybe under the hood it's not actually stronger and it works for some other reason, but practically speaking: if your style doesn't seem to work and uses compound words, try adding dashes.
I don't know if there's a way to directly tell the weight of the tokens, but there's some terms that are DEFINITELY more heavily weighter than others.
no suno, but midjourney messed up my subscription where I wasn't being charged but I also wasn't getting my credits for GPU time renewed for 3 months. And that was when they had only 1 guy hired for managing everyone's subscriptions.
I surmised as much, Midjourney behaves similarly, reinforcing your ideas with repetitiveness. They do have weights so you can control so much more, but that's an image gen, way different I think from here.
yes I have done that and can get into the prompt box but when I type it triggers one of the ideas suggestions and I can't escape out of it
it's a bug on the page
I hope they make some kind of weight system.
I feel like finding the balance between styles is a credit sink.
before today I used it in both simple and custom mode
especially when you are trying to combine two or 3 radically different musical styles together,
I'm not yet sure how to implement timing changes, nor do I even understand timing?
I'm a visual artist, I only dabbled in guitar and the piano lessons I did take were 32 years ago, so that info is essentially gone from my mind.
are you using Sonoma?
When I was personally getting familiar with Suno, I stuck to one style (indie math rock) and played around with a lot of different variations by adding a single extra prompt or metatag, and comparing the differences.
It takes a while, but after a bit you start to get a feel for what will overpower what, and while Suno still does whatever it wants, you have a bit more feel for how to guide it.
Near the end of your credit cycle, go wild with experiments to use the remaining balance before it refreshes -- or request different Prompt Experiment Playlists for me to blow my credits on over at the pool party > #1256301881966198805 -- dear lord, I don't know what to do with these credits, give me ideas to science them with
that's what im doing right now since it renews tomorrow.
burlesque math metal is pretty interesting.
I wouldn't mind using some of your credits. Hahahaha.xD
also
burlesque french math folk, orchestral classical metal, slam polka
wild weird stuff.
lmao I love it, guaranteed to give you a wild song, but I wonder how many of those are being "respected" given how strongly some of them are weighted -- unless there's one of those that you're more familiar with then the others, it could be hard to know exactly what you did.
yes, you are right, it's not selectable, let me try the upload audio button trick
Is there a guide maybe more of an unofficial one of brackets that actually work? I've already read what was on the official website and while that's decent I'm sure there are more.
Has anyone tried and succeeded with time signature prompts? Like 3:3, 3:4 or 6:8?
it won't work if you put it in instrumental.
"lyrics"
[D-Beats drums]
[buildup synthetic pianos]
[metal bass line drop]
[beat drop]
[electric theramin and citar chorus]
[galloping metal bridge]
[chugging guitars]
[epic piano overture]
Style :
cabaret metal, black metal, avante gard country ballad
also, I don't know what I am doing either.
but this makes face melting piano/guitar solos
@shut moat
I find it best to imagine anything you say to Suno as a suggestion rather than a command/instruction.
When it reads those words, it will use them as suggestions to give you types of music that include similar aspects, and it's not uncommon for it to "further push" a song into an existing "musical momentum", but personally I see it more as a success when they work, opposed to a failure when they don't.
true
Hi community, I've been trying out the free version of SUNO for less than a month. However, the latest songs are only generated with a female voice, despite my instructions to generate them with a male voice. Has anyone else experienced this, and do you know if it's resolved with the paid version? Thanks!
Absolutely zero difference between the models themselves between paid and free versions, only the amount of credits / concurrent songs / early-access to some features.
What different metatags or styles have you used to try to get male vocals?
ok I got that working, and I created a song successfully, thank you so much for working that out
how did you work it out, you must be a genius if you have no experience with Voiceover
I would never have gotten there without your help
nice, well done, I owe you big time
for the first time after thousand of songs, Suno is not capable of using the same voice when extending. Did it happen to you too or somebody just changed something at the model?
thank you, so much
Okay, thanks. I've tried using the metatags "male vocal," "male voice," and "male vocals."
try it in brackets when making the lyrics.
It seems like it might be having a hard time directly understanding "male voice", but you can try to describe the same thing in different ways -- "baritone" "growly" "strong" "broadly" "tenor"
you can alternate between female and male vocals I have found, but I think this might only be in custom mode which is in the paid version.
Custom mode is free as well
oh reeeeeleeee?? will I burnt all me credits in about 5 minutes.
then bought yearly immediately after
Just took this picture from a free account.
Okay, thank you very much. I'll try using those metatags.
I see a lot "clear vocals" from very experiment suno users. I've never felt "clear" was actually helping with voices. Should I rethink this? Probably haven't even tried in in V3.5, so maybe it could be different i suppose.
I think itโs the same mentality as using โbest quality, 4K UHDโ in SD prompts, imo.
I havenโt looked into it extensively, but Iโve yet to see something that could be used reliably
Worth trying anything. "Myspace" is all the rage today, and it is pretty distinct style. I actually just tested GeoCities, Facebook, Tumblr, and a few others and some of them did more nothing, but MySpace remains king, easily the strongest style among social network names ๐
I'd love to know how much of what we see in these types of prompts are influenced by people doing image or video prompting and trying to transfer those skills to Suno?
well, im back. i crashed. [METRONOME: 1 - 240bpm] works in conjunction with the genera i was not able to adjust the tempo manually the first couple were glitches.
but the prompt can aid in some cases
In Meta's Audiocraft/MusicGen, they trained on stock music sites, using the tags also. Some users tags songs with 320Kbps and so prompting stuff like that did increase quality. Of course MusicGen can't output any 320kbps, but that tag was not in all the songs, only a slice of them. and for whatever reason it ended up being associated with better music generally. i guess the laziest users didn't bother uploading in higher quality formats.
I haven't noticed it having any actual effect, same with "dolby atmos" and stuff like that. By far the best way I've found to improve vocal quality is to download a clip of the vocal, layer it with some bright synth sound in a DAW (playing the same melody), reimport it and input the lyrics again and extend. Obviously not completely reliable and means you have to start with that sample, but I had one of the most obnoxiously noisy vocals from v3.5 early access that I cleaned up by doing this.
It'll never be 100% perfectly clear though. It'll only be as good as Suno can get it.
i saw this floating around earlier, im not at a place where i can currently mess around with suno but im rlly intrigued - what does this style actually ... do?
#1256301881966198805 message Do you have any suggestions for metatags/styles for me to try? I wasn't able to get 200bpm in any of my attempts.
well, im back. i crashed. [METRONOME: 1 - 240bpm] works in conjunction with the genera i was not able to adjust the tempo manually the first couple were glitches.
[11:52 AM]@stratuss: but the prompt can aid in some cases
It feels like it draws out some more "amateur" content. Like, the kind of stuff people might record and upload to a social media site before it became all infested with viral videos etc. One person got some really bad singing, I managed to make something that sounded like a teenager talking... but it gives weird stuff sometimes as well like I got a loud scream
Yeah, I tried [metronome: 190] in that playlist I showed and it didn't work. I tried it a lot. Can you look at the playlist and let me know what I should have typed instead?
i tried math too to see it would work. 12 bars & 240 bpms
I named the songs what the resulting bpm was -- there was a 180 and a 175, which I suppose is a neat coincidence?
Does Metronome 4/4 do anything in the style prompt?
It's kind of a raw sound, or punk, more like an indie band without a studio. I didn't find it to be a unique style, actually, you can get that vibe with other prompts. (I basically specialize in prompting bad singers in Suno)
But it was unusually strong as a style tag compared to most "random" words when used as a style of music.
ohhh thats quite cool actually, i might try mixing that with some indie pop / bedroom type styles then! ty @unique stone + @untold briar , fingers crossed i get some cool results later ๐ซก
here yuh guys go <<<< try these [TEMPO: Fast, upbeat]
[METRONOME: 80-158bpm]
[TONIC: C] <<<< very important look it up if you dont know what a tonic is
[TUNE: EADGBE] <<<< tunes guitar
[TUNING: A440] <<<<< 440hz but this format seems to work too
Okay, I'm going to add these to the lyrics section, yes?
is this okay?
There's no spaces after any of the lines
you need < [TEMPO: Fast, upbeat]
[METRONOME: 80-158bpm]
[TONIC: C]
[TUNE: EADGBE]
[TUNING: A440] >
the little arrows
no and one space between coomas and next line
looks like it only added 4 characters, the angled brackets and the single-spaced padding
That's what we need?
and one space between [ ] []
That's a pretty wild gap in bpm -- are we able to be any more specific?
the metronome bit seems completely pointless with that wide of a range, like its probably gonna do smthn in that range anyway
i also dont understand necessarily why you'd need to specify the tuning of the guitar (if youre making a song with guitar in) considering its going to pick the right notes that fit your key anyway
BTW, while MySpace meaning "amateur music" fits so perfectly it feels like it can't possible just be a coincidence, I'd guess it is a coincidence. I prompt a lot of random words, and many things do something as a style.
I just tried Tumblr right now, and it's a weaker tag than MySpace. Kind of a vague easy listening or background music? https://suno.com/song/6cb2b96f-cc80-4661-be65-dfbda8744a51 I don't associate that style with "Tumblr" so it just feels random in this case, but if it had some music more tumblr-like (what would that be...), it would be easy to think "Of course it sounds like that"
Actually I changed my mind on MySpace being just a coincidence. Nothing about licensing data. Simple and obvious, I must be short on coffee today.
I have one stupid question. is there a difference in vocal quality between the paid and free version?
Nope ๐
oh yeah the other day i was just doing keyboard spam as the style, about three tags per generation, and i got some pretty incredible results and reusing the same styles of ... wordvomit was pretty cohesive which was rlly interesting!
i was afraid that would be the answer ๐
some songs i created with a free account are great, but vocals sound like 96Kbps mp3 ๐ฆ
https://suno.com/playlist/01fbf056-ba18-4d21-ac34-b69cc913eae2 I just added 8 more songs, targetting 20-40 bpm and got an average of 153 across 8 songs.
try elevis
Care to look at the most recent entries to let me know what I could have done differently?
This issue is that I didn't have "elevis" in the style section?
Okay, I want to make sure that I follow your instructions perfectly so I can help others learn how to do this.
The genre with the highest Beats Per Minute (BPM) is typically Speedcore, which is a form of electronic music characterized by a high tempo and aggressive themes1. Songs in this genre are usually classified as Speedcore at around 300+ BPM1. However, thereโs a sub-genre of Speedcore known as Extratone which has songs with a BPM of 1000 or higher1. At this tempo, the beats happen so fast that they sound like one constant note with a shifting pitch1.
Please note that these are extreme examples and most music genres have much lower BPMs. For instance, genres like Dubstep typically have BPMs around 135-145, while Drum and Bass averages a BPM of 160-1802. I hope this helps! Let me know if you have any other questions. ๐
Please let me know if I'm missing anything or need to retype anything
I have many questions, but let's start with the ones I asked first, and then we can move on once we figure that out โค๏ธ
np
I would be really excited to know how to prompt a specific BPM reliably -- other companies are specifically developing technologies to solve this, so it would be VERY exciting if we've figured out a way to solve the issue with our words alone!
i still dont understand why you'd prompt for a 'rough' tempo tho if the styles already are going to influence the suitable tempo that youre otherwise prompting for (and are aware of) ... unless you could be like 'Jungle but its at 120bpm' it feels like you can only already choose a suitable tempo the AI would pick anyway? like i wouldnt need to type a tempo range at all therefore
genres, I understand -- so we aren't able to specifically ask for a bpm or the tuning?
@charred merlin have you tried tempo tags?
tuning yes
So we are able to choose the tuning, but we aren't able to choose a specific BPM, correct?
im 85% yes on that
Okay, sorry, I must have misunderstood - earlier it seemed as if you were claiming it was possible to specifically choose the BPM by using [metronome: XX]
I haven't seen any examples that prove that you're able to do so reliably, maybe we'll have to make a #๐กโsuggestions-web to let them know we'd like that ability, since it doesn't seem to be promptable.
:\ Yeah -- I'm still experimenting with italian music terminology because I figured it would be strongly weighted, but there's inconsistent results.
But seriously, I need to get "Sunovator"pins made up to mail to people when they share ideas so good that I want to kiss them on the forehead.
I was waiting until I found some that felt like "superpowers" to share, but before I've reached that I've found that some work better in some instances than other -- age old rule of just reapproaching the issue and trying to find different ways to explain the same thing
combining genres can help to achieve the desired bpm, but not always ๐ฆ
I saw Yolkhead sharing some ideas about prompting in different languages, which really helps
Oh, there's ways to "nudge" a song toward specific tempos, but granular control over beats per minute just simply isn't there, from what I've seen.
I used largo to try to "calm down" my indie math-rock, but I didn't find any major differences from thematic prompt decorations like "sleepy" or "dreaming"
[entergetic] <<<<< helps
i got decent results with "High BPM" and it works in most cases, but there is no real control
i put it just after main genere(s)
it would be best if we had a separate "desired bpm" box ๐
like when you open fl and set bpm ๐
Lots of thematic prompts or "pseudocode" can nudge it in the right direction, but in my opinion the damage comes from people throwing whatever into the metatags/styles, being satisfied with the results, and believing that their input gave them 100% control over the output --it's fine if they enjoy the way it sounds, I'm just more worried about people who learn solely by copying and then wonder why it doesn't work for them.
definitely one of the top requests, esp for ppl taking their music making a bit more seriously. i always like what harber says about prompts and style tags being more of a suggestion than a hard rule so having more control over certain things that would be trainable would be great
And when the original artists are asked, the only explanation they give is "I dunno man, worked for me. Look at my prompts" -- just disregard all of these unsuccessful gens that I've swept under the rug
this is my little mockup idea of a future gen screen. if it was trained right something as simple as this would be handy! ;)
Love it! Have you posted that in #๐งโsuggestions-layout? I'm still talking about it here because it ties into the conversation, I just want to make sure you get as many discerning eyes on it as you can
๐ง oo good idea will do that in a min :)
Whether or not it's used in that fashion, stuff like this lets the devs know how passionately people feel about specific use cases -- it's amazing enough that people who are just breaking into the music scene are able to create this kind of music, but it's clear that there's a whole separate layer of musicians who want to find a way to squeeze more specific sounds out of this wild new instrument.
hmm...
i just opened this one in virtual dj, and it is 170bpm
I had 160 in prompt
eurodance, high energy, 160 bpm, hard kick, synths song. Listen and make your own with Suno.
The testing part comes in how reproducible the result is -- are you able to use the same prompt to reliably and consistently get the desired BPM? Or are you perhaps falling victim to the Barnum Effect? https://en.wikipedia.org/wiki/Barnum_effect Not trying to be dismissive, it's a genuine question that I regularly ask myself when testing anything.
The Barnum effect, also called the Forer effect or, less commonly, the BarnumโForer effect, is a common psychological phenomenon whereby individuals give high accuracy ratings to descriptions of their personality that supposedly are tailored specifically to them, yet which are in fact vague and general enough to apply to a wide range of people. ...
"Am I giving equal attention to the failed experiments as I do to the results that agree with my original theory?"
"Am I attributing personal meaning or relevance to vague or general results that could apply to many situations?"
The word "bpm" and even a 3 digit number probably do influence the style. Probably higher energy electronic music is more likely to be described in such terms. So using them may help you get he music you want. But I bet you try "80 bpm", "90 bpm", "100 bpm", up to 200, most of the songs will be in the same range of BPM no matter the number.
Well, maybe some effect, but rather than specific it might be "XX bpm" is slower than "XXX bpm". two digits instead of three. that kind of thing
it can be a coincidence, and you are probably right
i wish there is a way, but i haven't find it yet
Agreed, it certainly has SOME influence, as when people use "bpm" a lot, I've seen it cause weird results where the song increasingly speeds up, which would make sense if it's just associating it was "bpm == fast song, yeah?"
i have definitely got fairly reliable results before from prompting 120bpm with club music and house music, purely bcos i was using other descriptors and moods alongside the genres and was trying to get it to stay on track, but yeah if i try to ask for 90 bpm pop, i might end up with something at 145, likely just bcos of association and data knowledge
the rest of the prompt is definitely going to have influence on it too which is why it seems really unlikely that a metatag for a specific tempo will 100% hit it even if it does somewhat steer it in the right direction
There's something tangible there, so it's not like it's a useless prompt, it's just knowing how to use "the spell" as it ACTUALLY works, instead of what we imagine it to work as.
it reminds me a bit like teaching a baby how to walk and talk except im teaching... myself how to walk and talk so i can be understood by it ๐
It's definitely a humbling experience to realize that even the people making the product are learning as they go ๐
i making a stupi play list of the same settings same song same prompts 12.
I almost added a line "I wonder if the most common bpm numbers might work better... " Is 120 very common, like a standard-ish number?
for the club / house kind of styles, thats pretty much your bread and butter so it would defo be a big influence for those sorts of styles especially
i think most DAWs use 120 as a default starting point before you change it as well?
and it definitely understands tempo inference; i asked once for specifically edm progressive house which is 128bpm 99% of the time and about 16 straight generations i did all ended up at 128 (without me trying to specify a tempo at all)
How can I create 2 minute rock song with an insane number of chord changes? I tried this: "Create an intense hard rock song with 35 or more chords. Go for it!" but it just created a regular rock song.
Thanks for your advice. ๐
lol
A song. Listen and make your own with Suno.
sry its under 2 mins
lol
3 extends 12 hrs of learning about music and software prompts research 90$ worth of credits las 4 days. (not bragging serious)
hello, can we make multiple selections when clearing the trash
Yes! Click a song and ctrl + a to select all (move to trash or playlist)
lol
Click a song, and shift + click to select everything in between, or hold control and click to add or remove songs from selection
Can I ask what you're attempting to demonstrate with this, or were you just excited to share your song?
I feel the same way, so I'm wanting to know why I should listen to it, or how it's relevant?
Otherwise #๐ฃโsuno-showcase is a great place to share your songs
does metatag part in the beginning work for you, i mean normally it gives permsiison to only 120 character to describe the features of the song
i mean it is very long and specific
most times yes, i d share another total diffrent song to show you the contast
mind if dm you i dont want to annoyy any one
its ok
got you bro
"struggles to follow key signatures and BPM, and the songs don't always sound perfectly mixed and mastered" "prone to hallucination"
I'd say they're pretty straightforward, even addressed the creative liberties in song structure, like when Suno decides to try to fit 2 extra lines into a verse and messes up the entire rhyming scheme.
Then v3.5 came out, shoved that problem into a locker and took its lunch money. Which is a sign that v3.5 is still a work in progress, but therapy is doing wonders.
I'm not sure specifically how they refined v3.5, but it's so adamant on finding resolution or musical cohesion that I feel some people misunderstand what it's attempting to do -- given the context that it's extending from it can easily "fall into the groove" and exhibit the same problems that any Dad-Band rocking out in a garage encounters, they just have no friggen idea where to take the song, so they just keep riffing on "what would sound cool next". It's usually right, but frustrating when we want a nice clean end to the song that sounds musically complete.
The opposite issue is when you're extending from something, and to Suno's "ears", it feels like the perfect place to end a song -- regardless of how you want the song to go, it's completely convinced THAT'S the perfect place to wrap the song up, because it's musically satisfying based on all of the strict rules it's been given.
Without strong use of prompting (which can be either pure luck, or an intuitive feel for which prompts are more weighted than others), or a strong musical context for it to "hook onto" as an extension, v3.5 really leans into "generic" rock/rap/metal/etc
was there today some algorithm update? cause every extended uploaded audio comes out with sooooo much bleeding , after at least 20 tries i gave up . acoustic only .
As far as I know, the model only really changes if they say something about it in #๐โupdates or #๐ขโannouncements.
well , the outcome today was totally unusable ..
I've been getting a lot of really good results today, not to be dismissive of the results that you were getting. Can you share an example of a particularly bad generation so we could look at it?
When you write your own lyrics, it's remarkable how v3.5 is able to get them exactly like you pictured in your head. That's at least been the case for me, and it still blows me away. When it first came out, I extended all of my favourite v3 songs with "quirky flows" and it was just a night and day change. (except I'd sort of grown fond on the quirks ๐ ๐ )
https://suno.com/song/fb275612-4490-4b35-b1f6-239a82e44255 after 0.33 its really messed up
, deep,phonk song. Listen and make your own with Suno.
https://suno.com/song/f34b0bad-100a-4339-ba6f-ee8e190e5115 based on this clip from my daw
A song. Listen and make your own with Suno.
No, I just meant how I would write a regular stanza of lyrics and it would interpret line breaks awkwardly and try to fit 2 lines of lyrics into a spot where there should have only been 1. They didn't sound objectively horrible, but the timings were always ...off.
I know my opinion is subjective, but I don't think it sounds bad? Can you explain what you were hoping for vs what you're noticing?
yep
tbf, if I was at a club for some reason and this song came on, I wouldn't be complaining to the people who dragged me there about the music.
its a test anyway
too much bleeding .. its too much mixed up . i know its IA and no wonder machine .. but sometimes frustrating lol
my ears are complaining lol
#๐ฌโgeneral-chat message
This issue with DistroLock rejection is also happening with me. Does Suno fingerprint it's own creations using AcoustID or similar services in order to enforce copyright protections? Not really sure who to address about this. I'm experiencing this problem with DistroLock and new Suno songs that haven't been made public or distributed.
/chirp
Chirp is retired, and no longer on discord. You can make music now on our website: https://suno.com
ๆ ๆๆฒปๆ้ณไน/chirp
i took out hall and delay .. results are a bit better .. tx
๐ค๐
https://www.suno.com to make music โค๏ธ
Are you uploading songs made on a free account?
Iโve had no issues with DistroKid or the DistroLock service, but all of my songs were original generations and extensions
That being said, DistroKid support themselves is probably the most direct route for answers regarding them rejecting songs.
Hi! Can you generate Suno songs in Discord?
Both Suno and DistroKid accounts I subscribe to are paid (DistroKid through FL Cloud). I've tried to get in touch with DistroKid and I swear they must be using chat bots. They simply reiterate what you say and send you on your way. It's been a bit frustrating customer service wise.
Besides this, I'm really just wondering if Suno uses fingerprinting services for audio created on their platform. Not only do I get DistroLock rejections, I also can't reupload an idea created by Suno after working with it in FL Studio. These reuploads to Suno also get flagged as existing copyrighted work.
i've gotten the audio base thing as well after working on it in FL as well
/chirp
Chirp is retired, and no longer on discord. You can make music now on our website: https://suno.com
Yeah this copyright stuff is proving to be real tricky.
i use flstudio and distrokid , and so far ive no problems . but all song text are mine , album covers too
Not any more, I just replied to someone just above trying to use the chirp command
I see, thanks
Edit: no problems with my own work either. Only experience this problem at times with Suno generated audio.
i used suno audio but processed via the daw
Yeah, its been kinda hit or miss for me
even after processing, or just recreating the melody Suno made, using different instrumentation
and rendered wav at 24 bit wav
Hey, everyone. Another quick series of questions, because I am too bloody-minded to give up on this song. Is there a PERSONAL component to how the AI algorithm works? Like, in addition to the base training data, does it learn from interacting with you specifically (figuring out what you mean by/want from prompts, etc)? And if so, is there a way to purge/reset that data? Some people have been trying to help me with my current project, and for some reason, the past week, Suno has been getting more and more delusional and less and less useable for me, and I'm starting to suspect I may somehow be the problem. ๐ฆ
There is no personal component, nothing to reset. It can feel that way, but probably you have a tricky prompt. I bet if you try a totally new song and style, back to normal.
Okay, let me give the AI a break from trying to figure this out, then. Thanks... It helps to know that I didn't somehow pollute my own ability to make music here.
Played a bit more with the 'token' concept in prompts. I don't think tokenization matters.
The obversations are true. Small text tweaks (dash instead of space, capital instead of lowercase) can have very big effects. But how the text is tokenized doesn't seem to matter at all.
Invented compound words in styles might work or not work, but using tokens for the pieces didn't work more often not. Such a fun concept, Suno should make styles-as-lego-bricks into a prompting feature. If anyone has a clear test case of style chunks based on 'tokens' @ me. For now I'm a token skeptic. @alpine venture
Hey, everyone. Another quick question. If I upgrade my membership prior to the next billing date, does it A ) take effect next cycle, B ) bill me fully for the "new membership" now and give me all the benefits, or C ) bill me the difference for my membership tiers, and give me the difference in tokens? Also, @untold briar Do you mind explaining a bit more about what you mean by tokenization? Kind of curious; I've heard the word bandied about before, but have no idea what it's meant to encapsulate.
Capitals can matter, so it's not implausible that is good advice. But Thinking about it now, my guess is COUNTRY is not different than country. When capitals do matter, in my experience it's on the more unusual prompts, more edge cases and unpredictable things.
I've looked at that plenty of times. I'm looking for more real world usage where people have tried things and it has worked. Especially the intros but I would like to have something that's a little more concrete.
Text formatting in general can have big effects. Like you can prompt a duet with text format by itself, at least sometimes, just with:
PERSON: Hello I am person
PERSON: I am also a person
#1252267444773326848 is the closest that we have on the Discord, those are user-submitted tips. They might not be outright formatted as lists of metatags, because as I usually say, that's like asking for a list of things you can say to ChatGPT -- people have shared their experiences in the workflow section.
I have some Prompt Experiment Playlists over at #1256301881966198805, and I'm open to taking requests if there's something you'd like me to spend my own credits on to test for you โค๏ธ
billing@suno.com would be the best place to ask for account & billing related stuff โค๏ธ
Asking for what others are trying is not asking to be sent back to the official website. There has to be other websites around that people have posted about their experiences what has worked what has done different things. That's what I'm asking about. I don't want to go back and look at the same materials I've been looking at for months. That's why I'm here to get better support.
The wiki is the community-sourced website -- if there are others, I'd love for you to show them to us.
I'm not sending you to the official website. The official website is https://suno-ai.notion.site/FAQs-b72601b96de44e5cacd2cd6baa985448
#1252267444773326848 isn't the official website, it's the guides & suggestions curated by the active members of the discord community.
imo ... the absolute most bestest support for using suno... is just experimenting and seeing what you find works (and what doesnt!), so you can learn and improve on it. there isnt an overall one size fits all really... plus its a lot more fun this way anyway! but the community wiki can defo be helpful. ๐ just dont believe that adding 60 words in square brackets will instantly solve all of your issues ;)
If you have specific questions, let us know how we can help you try to make the music you want to make -- we're here to help collaboratively, but there's no "magic guide" we can point you to.
There's different discord channels that other users have started to discuss tips, maybe you could go into #๐ฌโgeneral-chat and ask for people who know the answers to DM you. I'm sorry that the community wiki and user-submitted content weren't sufficiently helpful ๐
I have no idea what you're basing that assertion on -- I'm simply saying I personally don't know of any other website, and I'm sharing the content that I'm aware of.
Has someone told you that you're not allowed to share other websites? That would seem silly.
Hi! I haven't done a search yet, but has anyone figured out how to insert sound effects? I am trying to add police sirens to a part, and he literally sings POLICE SIRENS! ahahhaha. Another song said HORN SOLO then the words!
I've seen sirens accidentally generated more than I've seen them intentionally prompted ๐
Is the suno website down?
I'd love to tete-a-tete about this some time ๐ Now you got me questioning my own confirmation bias and I'm squirming in my seat, damn you!
Jokes aside, this would be a good example for another prompt experiment ๐ฌ ๐ช
The only thing I know for sure about capitals is they can change things... sometimes. Most often I think they don't change much. Capitalizing words in the that lyrics that are meant to be sung does work as people probably expect, the singer more likely to emphasize or even yell that word.
for like 6 songs in a row, it's inserting the style of music INTO the musical lyrics before the song
anyone else have this happen?
for those of you who are using distrokid to upload to spotify, are you doing full albums or just singles? I have 22 songs but I feel that might be too many for one release.
You can do up to 30-someodd at once for an album -- the only difference is you'll have more songs to enter the information for at once.
I have a 3-song EP, and a 12-song album.
What were you seeing when you thought it was working? Sometimes parts of style tags work as tags alone, or combined with other words. But using the exact tokens did not seem to matter. Sometimes works, sometimes doesn't.
is v4 coming soon seems likes way overdo or is that lawsuit slowing you guys down? @unique dome @graceful cliff
Like I said, you're making my question my own confirmation bias, as I'd always believed there were differences in variations of indie(-)math, and had inferred that dance(-)punk would have similar differences, so I'm making a bigger comparative playlist to see what I can dredge up.
I respect your input on the matter, and if you're willing to throw down bold text about it, I'm paying attention. ๐
My immediate takeaways are that I would classify Danko Jones as dance-punk if Suno knows what it's talking about.
I didn't mean to suggest that wouldn't work. Just that looking at the tokenizer breakdown of that text doesn't let you predict if it works. Dashes have big effects. Parenthesis have big effects. I'd be a little surprised if a dash in parenthesis changed nothing, suno seems to care a lot about those characters.
๐ I didn't take it as anything negative! If anything, it's good to be pushed out of my comfort zone a little bit, because now I'm really exploring my own suggestions from the perspective of someone who might be trying a whole different style than me. My use case of the tokenizer was simply for checking if hyphenating something "bridged" the words, or if the hyphen suffixed or prefixed itself to a token (as is does in the case of "indie math-rock", but I just prefer the way it looks, so call me superstitious ๐ )
I really enjoyed seeing the research yourself and @alpine venture were diving into, and so the fact that you'd been looking so closely at things lately made me want to re-examine my own methods. Early conclusions are just re-affirming that v3 is willing to get a little more diverse, so I might need to refocus my efforts on comparing hyphenated / non-hyphenated prompting on just that model before bringing v3.5 into the mix.
I'm scurrying off to bed while I can keep my eyes open long enough to make my way there, though -- I'll give you a ping if I find anything interesting one way or the other, Jon โค๏ธ
My use case of the tokenizer was simply for checking if hyphenating something "bridged" the words, or if the hyphen suffixed or prefixed itself to a token (as is does in the case of "indie math-rock", but I just prefer the way it looks, so call me superstitious ๐ )
If bridging or prefixing in the token determined whether hyphens effected the music or not, that totally would prove tokens matter. But your example is a little more complicated and not as simple to test as what I tried.
having not subscribed in a while, it's great to be back to experience suno's passive aggressive nature ... I asked for male vocals, and for 6 clips, it gave me female, I added "fe" to the male, leaving the rest of the prompt unchanged, and I immediately got 2 male vocals. I think the suno developer's should be very careful about this hateful creature they're creating becaues it is clear that it hates us, the overly upbeat and sentimental song default nothwithstanding. -- Either that, or there is some sort of "contrary" setting in the UI that causes the bot to eat up credits. Too close to call.
/chirp
Chirp is retired, and no longer on discord. You can make music now on our website: https://suno.com
Are they coming up with new vocalists on suno? They all sound the same
what are producer tags? i tried to use the words the danger, the danger in my song and it won't let me, says it's a producer tag?
Producer tags are like those short shoutouts at the beginning of the music, like "We The Best Music"
When will the interface on mobile work properly? I cannot download wav files and changing to desktop on such a small display is a real pita --.
Also in addition to that, guess what.
Desktop is sometimes broken too --
Guys is there a way to find the tempo of the song from suno and also is it possible can we use the same seed to generate another version or updated version of the song with the same seed
Is there a link to the comprehensive list of tags like [Verse] you can use? I saw somewhere someone mentioning instructions for song structure to the measure bar level and used ##
tokenizer doesn't work like this. it doesn't matter if you add spaces or not. It least it should not. But you should be able tu use {} too. Should not matter. Tokenizer and GPT is not programming language, it is not so exact, does not work like that, it reads context.
experimental avant-garde glitch myspace, licensed fx, theatrical hypersilence, applause, office song. Listen and make your own with Suno.
p interesting
I remember not long ago it was almost impossible to make non-music with Suno
3.5
and getting harder
and over the last little bit it's gotten a lot easier
and to variate existing songs
whatever you guys are doing imo it worked well
I see a tendency to generic voices. But on the other hand I never had trouble creating a more unique sound. The only trouble I had on this regard is creating old voices. That actually never worked out for me.
How do you prompt for low male voices in multiple genres?
Suno leans SO HARD into high tenor and soprano, it's sooooooo frustrating
Direct prompting seems to just make it worse
this is some bs
https://suno.com/song/4b21afb5-325c-44c5-80b0-e95d60b4e443
look at the prompt and listen to the song
this is with 3.5
This sometimes makes me livid too. I was working on a song and I specifically wrote "female vocals" in the style prompt, early in the prompt too, and it literally felt like it slowly started morphing into a more masculine-sounding voice until it just sounded completely male in Verse 2. Or sometimes there were 2 singers (1 female singer) no matter what. It's irritating to say the least.
THIS
I start to just feel hopeless and get burnt out
It comes and goes in phases. Check the date on this one https://suno.com/song/c298dcdd-55d8-4351-9dec-d2b165526e61
In your style prompt, try replacing strong male lead with strong male vocals
I've tried a lot like that
For gender, the first thing to try is just repeat the gender in the style prompt like literally "male, male, male".
that's ridiculous, if that works. Ridiculous that it's needed
That's how I fixed that male thing in the song I mentioned. One moment I'll copy it
https://suno.com/song/e0ea53d4-b9e7-4368-8ec9-7831b85457e7
I put "female vocals, female" as the very first keywords
female vocals, female, Electronic, Dance, Dance-pop, Pop, EDM, House, Synth-pop, Euro house, Hi-nrg, Electro, Tech house song. Listen and make your own with Suno.
this song feels very disjointed, lol
but congrats on the voice at least, haha
It helps to know that prompts in image and audio generative AI work a bit like spreading butter. The farther you go, the less butter you spread, so the weight is the heaviest at the beginning
oh, yeah, but I have tons more success with midjourney
Could you elaborate on what you mean by "disjointed"? But in any case it's satire
and much less technical prompting
Oh, the instruments and the vocals feel like two different songs
One could argue this adds to the comedic effect
fair lol
One time I got surprisingly clear vocals. Almost udio-like in terms of vocal clarity (otherwise I have barely had success with Udio as opposed to Suno). But they were slightly off-sync. Extending from a certain point makes them sound more in-sync and it's easy to tell where the extension begins
house, dark, electro, female vocals, piano, edm, dance, emotional, sad, melancholic, clear vocals, bass song. Listen and make your own with Suno.
so far the only tried and true method I've gotten to work consistently for male vocals at least is to prompt with a voice clip of myself
Does Suno allow using your own voice???
Lyrical ballad, medieval bard, lute, whimsical, dark song. Listen and make your own with Suno.
I had my wrists slapped for uploading a song with vocals (which was generated with Udio and I wanted to extend with Suno)
just don't upload anything that already exists
So using the browser recorder bypasses that?
that's pretty odd
Wait let's get on the same page. Does Suno detect existing singers somehow? I thought the detection system was less convoluted, i.e. "vocals = no"
How come? I remember when this feature dropped that it didn't like vocals of any kind
you can't set as public, but you can still share links
Right.......
here's one that's literally how my voice sounds:
https://suno.com/song/c6c56489-06fc-4f3d-85ff-ec24954dd910
Low-key pop, chillwave, lofi, soft tones song. Listen and make your own with Suno.
like, uncanny close
I sang the first two lines, then wrote the same first two lines as the first part of the song, and generated the rest from them
That's good to know. I could experiment with that ๐ค
Just that it didn't click that the only penalty is not being able to publish on the website itself
ye
But can you download the mp3/wav file and upload it elsewhere?
yeah I think so
I guess it's not so penalizing then ๐ค
ye
I hope that Suno doesn't assume I'm female though (I'm an adult man but I'm a tenor)
It'll be hilarious
if you use that method I described above, should be good. I only used like a 7 second clip, longer would probably be better
If you only want the voice, even 7 seconds is fine
yeah
If you use longer clip, it can also copy a longer melody etc so songs you make from long uploads could be too similar. Though you can just choose Extend From time.
I managed to create a convincing hardbass song. Here's the tag. This song is an example.
Donk, russian hard bass, techno, hardstyle, pumping house, bounce music, bass boosted, 160 bpm
I used more than 60.000 credits to make music from at least different 12 genres in more than 7 languages and I did not notice at all that "they all sound the same".... on the contrary ! It is amazing how many different voices there are...the voices are the best thing at suno..
I think prompting in many languages probably helps increase the diversity of music and voices. The lyrics 100% influence the music, and you're prompting with much more diverse lyrics using 7 languages.
Please I need help with doing a track with male and female vocals on it๐คฒ
The tags don't work. It was a "miracle"
i have sent messages to Suno staff and also mails to the suno email account without any response...
I just want to know if it's possible to apply for a job with Suno while living outside the United States.
i have a question regarding credits and monthly refresh on credits date. I sub'd to annual on certain day. does my refresh of creds occur on same day as sub or on 1st of the month?
on deletion, since for now cant select multiple there r some issues to attempts at multi delete that would req a ui make over.
with making managing ur content maybe a little easier, especially with how theres limit on content per page, its best practice to work only one or handful of projects at time and delete the ones you dont want as you go. Alternatively, you can also efficiently optimize your profile content by going thru and deleting things at regular frequency. like 1x a month or 1x every 2 weeks. These can help you manage your content libraries more efficiently.
lastly theres the old school method, download/offport everything as you go and delete everything you dont want as final content on your profile. last option is more the nuclear mode, as removing content, you lose the ability to promote that removed content in platform network once its gone. so only remove stuff you dont want to use/dont need to share.
as to why multiple delete options would require UI make over
rn delete option is under ... menu per track/item created. the list itself is fetched in a sort of live state thing as well. for instance
if you sort your content library by 'top' and a song gets likes while you are viewing it, it changes position in the list. while you dont see that change unless you refresh the page, the delete option is pegged to the content object via the parent content specific submenu.
changing delete option to be say more email UI where its dependant to the list directory instead of parented to the content object, would be a slightly more involved overhaul. to make sure that everything is transitioned correctly with out code conflicts.
long story short there is ways to do it, but its kinda gonna need a big code overhaul to ensure it works properly.
Hereโs 3 ways to select multiple songs at once โค๏ธ
Monthly credits refresh at the same time you purchased them afaik โ subscribed at 6pm? Thatโs when credits refresh
Hello, I propose that "everyone" can change the intro of a song, because often, it happens that everything is perfect, except for a defect at the beginning
While we're hoping for "in-painting" in the upcoming model v4, this isn't something that we're able to manage at the moment, so I hope that you share your support for the idea over in #๐กโsuggestions-web ๐ช
Well, i have some difficulty, with hashes in my first message, deleted ... So i did it now! Thanks!
Woodhouse's diet consists of sugared grapefruit and any Discord messages that don't follow his secret list of rules ๐
Thanks for sharing the suggestion ๐
i meant more like day of month wise? if i subbed on say 5th, would my next months balance hit on 1st of month or 5th?
5th, sorry, I thought I typed that as well ๐ not enough coffee at that hour, I suppose
no worries thnx
Hey, it's me again^^ I have a question, not essential, but which bothers me a little, I will take my example, I manage all the Suno accounts of the family, with their respective email addresses of course, so, what are, if any, when it comes to managing IP addresses?
May I ask, how is it possible to cover Smells Like Teen Spirit from Nirvana?
Yo โค๏ธ
Iโm not sure specifically, but Iโd advise that it couldnโt hurt to be proactive and reach out to billing@suno.com (they handle account related stuff specifically, so it might be a quicker route than support@suno.com) โ depending on usage rates they might have specific guidance so it doesnโt look like a single user making use of multiple accounts.
Iโm not on the support team, so the advice should be taken with a grain of salt, but I always imagine that being upfront, honest and proactive can help alleviate any confusion โ most of the time, at least. ๐คทโโ๏ธ
you can't it is copyright
Technically you canโt use Suno to make a cover, as the lyrics are copyrighted and you canโt upload audio content you donโt own, but you could use reminiscent lyrics and similar style prompts to make a similar song or a parody song?
โAngsty-teenage-punk, palm-muted guitar, grungeโ
Ok, it's because I recently listened to a crazy good cover. Won't deepen that though because I don't want to give this guy trouble.
it wouldn't have been done in Suno, Suno will refuse to do copyright lyrics
Ok, thank you very much, sincerely, I don't think so either, because each account is separated, both by the address, and by the possible PayPal... So no fraud possible in theory...
Uhhh. Yeah. ๐
Suno doesn't always recognize copyrighted lyrics, though.
oh really?
Yeah. Still very against ToS to upload content you donโt own, but you can buy the rights to a cover license from the Harry Fox Agency for example
Yea. I don't know how often it misses them, but we'll occasionally see stuff posted to the Showcase channel that was extended from an upload of protected material
The Real Slim Shady slipped through a couple days ago.
would u link it, pls?
To the deleted post, or the banned user? 
Ahh, I understand โ shouldnโt be an issue if theyโre paid accounts, I would have worried more if it was multiple free accounts as not to seem like free credits were being taken advantage of.
Probably canโt hurt to reach out and clarify, but I canโt imagine there should be an issue ๐
Wooo ?!
Generations under 8 seconds are free โ pretty sweet when youโre attempting to end a song and donโt have to pay ๐
Likely is because of when you are extending from
Sometimes v3.5 gets stuck this way. Try v3
Make sure youโre not extending from silence
It's not silence, sure
v3 don't work
And i lost credit x)
v2 don't work too
I have only failures... over more than 30 attempts.
Some credit are not being repaid... I'm losing out.
Could you share the part 1 you are trying to extend from?
I think itโs an audio upload if you canโt use v3
That was what I was thinking it might be
Only generations under 8 seconds are refunded.
Like our hamster mascot kindly asked, can you share pt 1 that youโre extending from so we can try to help figure out what the issue might be?
Without that, all weโre able to suggest is that you extend from different timestamps or adjust the clip youโre uploading.
I send it
Okay, im out for a bit but ill give it a shot when i get home โค๏ธ
All my attempts are under 8 seconds, but I still lost credits.
That appears to be part of: https://www.youtube.com/watch?v=XoLouT7TqZY
My sample are a remix
I use fruity loops for this
Look
All refund, but, i lost credit
Why are you saying "we" are you working for them?
โWeโ as in โwe the community hope to see thisโ
Not on staff, just an evangelistic user ๐
...then why does it sound identical to that YouTube video?
It's not identical
That is not the problem right now...
0 credit and "all refund" x)
Let me demonstrate my concern.
Left channel is the YouTube video.
Right channel is the audio clip you sent me.
What's the source of the guitar part in the clip above?
" don't understand why my 'work' is being judged when that is not the issue right now. I'm using credits to generate a continuation for my remix and it's not working, and the credits are not being refunded.
This sample
and now ?
Do you want all of my work for a review, or can we address the real issue at hand?
is there a trick to get the AI to use the lyrics i provided instead of mashing up random phonemes that no larynx has ever produced?
You might just need to wait a couple minutes for the credit refunds to reflect on your account, as they all say refunded โ if youโre doing a lot at once, it might need some time to catch up and calculate?
Can you share an example where itโs particularly butchering your lyrics?
Iโve had audio hallucinations, but I canโt say Iโve experienced such a drastic issue, so Iโm not sure how best to advise.
Right, just F5 and it's ok ๐
Okay cool!
Definitely still frustrating when v3.5 gets โstuckโ like that, but at least we could shine a light on the credits. ๐๐ฆ
Is there a way finding playlists I've liked? Except over the search function of course, there will be a time I don't remember the name anymore.
mods and guides will always point out tos violations, because we dont want them here. ez
It's a remix that I made MYSELF. ><
putting a filter on something isnt a remix...
AND you still need the copyrights to do a remix in the first place. so...
doesnt make it better
It is written on his website that we have the right to use it if we credit him (which is why my remix on Suno is called MDK).
gets get something straight here : I dont care what you do or dont do. And i dont care if you get sued or not. I just care about sunos tos wich want you to have the full rights to a song. You dont have to provide anything to us. We just dont want content here that is potentially a tos violation and can cause trouble for the discord.
I know it's pretty strict wording, but the Audio Upload thing you agreed to says that we can only upload things that we own 100% -- I know using stuff with permission feels like it should be enough, and it usually all you need in most other places, but specifically for Suno they want stuff only that's 100% your own.
Not that it would contribute to the generations being cut short after a couple seconds, but just to toss my 2 cents into the mix โค๏ธ

You're 100% right, that gives you the right to that material, but it's "shared" ownership, instead of 100% exclusive ownership
So still granted permission, but it doesn't override Suno's rules
Ok
Like if you have an e-cigarette that technically doesn't break the "no smoking" rules, a store owner is still within their rights to say they don't want e-cigarettes because it still gives the appearance of smoking.
depending on the country
I owned a cafe when e-cigs became a thing, I had this conversation many times as the seniors who frequented my shop were asking why people were smoking inside.
In France, "no smoking" applies to everything, including e-cigarettes (but that's not the point).
Vivez-vous en France?
Oui
here's my brain spill for the record: basically, if youre getting stems for a remix, youre being provided with tools to create a derivative work from them, you arent getting ownership of them or exclusive rights to them which is the issue. i recently downloaded the stems for a david guetta remix competition which were super fun to have and work on but i definitely wouldnt want to upload those bcos even tho ive been given them to use, i dont own them
Now you're starting to understand how difficult the concept of content moderation is, and why it's easier to take a strict approach to it -- people have been attempting to re-use other people's music as their own for decades, and the concept of what changes constitute a legally distinct new work of art are a super complicated matter.
It wasn't for commercial purposes, just to create a different remix. I know that the music doesn't belong to me (which is why I mentioned MDK).
https://youtu.be/O3pyCGnZzYA A Tribe Called Quest's "Can I Kick it" hasn't earned a cent because they used the bassline from Lou Reed's "A Walk on the Wild Side" https://youtu.be/0KaWSOlASWc
Official HD Video for โCan I Kick It?โ by A Tribe Called Quest
Listen to A Tribe Called Quest: https://ATCQ.lnk.to/listenYD
Watch more videos by A Tribe Called Quest: https://ATCQ.lnk.to/listenYD/youtube
Subscribe to the official A Tribe Called Quest YouTube channel: https://ATCQ.lnk.to/subscribeYD
Follow A Tribe Called Quest
Facebook: https:/...
Rest in Peace to Holly [Woodlawn] who came from Miami, F-L-A, and who Lou first mentions in the opening lines of the song
Congratulations (a long overdue one at that) to Lou finally being inducted into the Rock & Roll Hall of Fame
*So sad to hear about Lou's passing today at the age of 71. There will never be another one like you Lou...
I've done moderation in several places and for both larger and smaller communities. I don't think it's right to speak poorly to someone without answering the question. A simple 'remixes are not allowed either' would have been sufficient, I think.
thats still covered by "I certify that I own or exclusively control all rights" which you don't if you've been given stems to someone else's song ...
I apologize if you felt that I was speaking to you poorly, I was just trying to highlight the difficulty of figuring out how the Suno platform specifically fits into the larger view of the music industry.
You're 100% correct that the original artist has given permission to remix his songs on other platforms if given credit -- but his permission isn't the final word -- if you ask your parents if you can go swim in the neighbour's pool, you still need to ask the neighbours first.
For the sake of clarity; In terms of Suno, you can't remix other people's songs.
The stems and samples I used for my remix were directly provided by MDK himself to the community. That's why it seemed legal to me.
I wasnโt talking about you when I mentioned speaking poorly.
Sorry, too much coffee and I'm feeling jumpy ๐
Admittedly, I don't envy the moderator's job, because on top of having to stop people from sharing dick-pics, they also want to make sure that they clearly state the rules -- not that they're trying to be downers, but there's legitimate legal consequences, both for the users AND potentially for Suno.
It's a really cool tool, and we're not even a year out yet, so I think they're focused on trying to stop a few people from openly encouraging others to "break rules" because it could cause unfortunate things to happen to a growing product that we're all really excited about.
Imagine if you REALLY loved a robot, and asked people not to push the robot over, and the ENTIRE day that you're watching it, you have people hovering their fingers over it going "I'm not touching it... I'm still not touching it..." -- then touching it quickly and running away.
๐ซจ

Creating music from the internet (Yes, there has to be a source for AI to create) is not without legal consequences? (Real question)
Ooooh, this is the exciting part! It doesn't NEED existing music to "reassemble" -- I explained a similar version of how you could make music out of tv static a couple of days ago, if you want me to track down that explanation, or if you'd want me to chat about it here?
Give the link to the explanation to avoid cluttering this channel.
(well, yes, it REALLY helps speed the process up by years, if not decades, by having data to train on, but there's places that sell training data so you can handle it the right way)
Heck, don't worry about cluttering up this channel, it's awesome to ask these kinds of questions because it helps us all learn.
100% guarantee there's someone out there who is thinking the exact same thing you are, but they couldn't be bothered to hang out with us and talk about it โค๏ธ
I know I tend to be really flower & colourful in my analogies, so let me know if there's anything I can re-explain ๐ช
Ok. But the robot that creates the music did use a database (whether Suno paid for it or not) of known music and royalty-free music (which might not be royalty-free anymore, by the way).
Like AI with images.
it's awesome to ask these kinds of questions because it helps us all learn.
i would 100% read a theharber ebook on all of this stuff and knowledge and experiments ๐
Tomorrow, I create a music track on Suno, and it unintentionally uses the lead from a known songโwhat happens?
Now, in terms of image gen, there's a project called LAION; Largescale Artificial Intelligence Open Network -- that provides an open-sourced, metadata-tagged set of 5 billion images for people to use when training image-based AI models. https://en.wikipedia.org/wiki/LAION
When I say the lead, that's just an example; it could be a bassline, a guitar line, etc., etc.
So, Suno hasn't disclosed the dataset -- admittedly, I don't KNOW if there's an audio version of LAION, I know that LANDR is making their Fair Trade AI program, but that's not free in the same way.
A lot of people are really eager to understand how Suno works, but I think that most explanations like "using a guitar hook" are a drastic oversimplification of the process, because it's not "re-using" content, it likely just inferred what kinds of patterns commonly arise within music so it could try to recreate similar patterns -- then with extensive RLHF (reinforced learning from human feedback) it can learn which generations fit the expected output and which didn't so it could refine itself for the next iteration.
I'm happy to discuss technical speculation, but as much as we can talk about it in polite company there's a pretty serious on-going conversation about the dataset. Most we can do at the moment is wait and let the conversations happen between people wearing suits costing more than our yearly rent.
loveyouharri
tbf with the phenomenal amount of data - and therefore noise - used for training models like these, the chances of it ending up recreating something 'unique' (eg a melody) is ridiculously slim and would be 100% a coincidence due to the fact the AI doesn't just target and reference a single section or song for particular parts or prompts
I'm not sure, but I may have made Suno's first crappy live concert.
I've experimenting with prompts for fully acted out scenes, feel like i'm onto something
https://suno.com/song/00af7fa2-de15-40e5-9f8a-faad56c95dea
Live Rock Concert song. Listen and make your own with Suno.
Has anyone used () bracket prompts within [] bracket prompts before?
are Suno WAV files just Suno MP3 files converted to WAV because they don't seem to be any better quality
from what ive experienced, i believe so; it seems to be there for file conversion purposes (eg to let someone immediately have a wav format of it) rather than for better quality or source audio (bcos the source seems to only be as good as the mp3 provides) but it does make me wonder whether in the backend, suno composes a 'full' version before the mp3 compression kicks in and we just can't access it.
It's pretty poor form, Suno make it seem like it's going to be a better quality file but it totally isn't
Stems which aren't stems, WAV files which are just converted MP3... Suno needs to be more honest and transparent
Hi Guys,
Question:
Can I put everything I need in the Style of Music field with meta-tags
to create an instrumental song?
From genre and style, BPM and instruments?
Like this on lyrics:
[instrumental]
[genre: DEEP HOUSE]
[style:arabic]
[tempo:90]
[intro:violin solo]
[chorus:violin, electronic drums]
Suno does not, and can not, be controlled like that, no. You can build instrumentals in the lyrics box with metatags though
It's all ~vibes~
Thanks KakerMix...
The style of music field is too small to work with several meta tags and have greater control.
It's what the pros do, yeah
Lots of people do it different ways, but I like to use
{}{}{}{}{}{}{}
or
!~~~~!!
as stand-ins for lyrics
Something that Suno can see but can't sing
Nah it's just that the audio quality produced by the model isn't much higher than the MP3 bitrate they offer. When MP3 was the only thing available people kept asking for WAV, the devs kept saying "but it won't sound better because of the model limitations" but people kept asking anyway.
There are some higher frequencies present in the WAV that are missing from the MP3 so it definitely isn't just a conversion from MP3, it's just that there isn't much to hear. This might change with future models as quality improves.
There was a short time after WAVs became available that there was a bug that caused it to just download the MP3, renamed to WAV. That was silly.
I wonder what effect the watermarking has on quality
Supposedly it's inaudible. I have no idea how that works thoughh.
I believe it will be a slow data format like WSPR or FT8 in radio, which probably contains a marker and a pointer to the creator
Do we have updated documentation about prompts for version 3.5?
Could you send me the link?
there is a Wiki but I don't know how to get the link
have a look at the pinned messages
Can someone explain
what this is
!~~~!~!
I mean, they explained it right after that example
Also see the percussion example near the bottom of https://www.suno.wiki/faq/metatags/instrumental-tags/
It's a way to give a song structure without singing if you are doing an instrumental but not using the instrumental mode
Like this:
[Verse 1]
!~!~!~!~!
!
!~!~!~!~!
!
[Verse 2]
{}{}{}{}{}
{}{}{}{}{}
{}{}{}{}{}
{}{}{}{}{}
[Verse 3]
{}{}{}{}{}
{}{}{}{}{}
{}{}{}{}{}
{}{}{}{}{}
[chorus]
{}{}{}{}{}
{}{}{}{}{}
{}{}{}{}{}
{}{}{}{}{}
That's how I use it
Everything with Suno is a suggestion, but it just gives you structure without lyrics
thanks, but what is . and what is ! in this example?
[Percussion Break]
. .! .. .!
!! ... ! ! !
same for
!~
and
{}
i understand the concept, but not "execution"
silent words basically, if I'm following what Kaker is putting down, suno wont actually say them as lyrics sung, but will still give the structure of them being there, so sections wont be made shorter than they should be.
Can someone help me?. I have been messing around with my old band extending etc, but i wanted to know if there was a way i could upload the 60 secs of audio and get the AI to just enhance what i have given it.
as i wrote, i understand the concept, but i have no idea what this should do
[Percussion Break]
. .! .. .!
!! ... ! ! !
what is dot here? bass drum, snare?
what is an exclamation mark here? snare, clap, crash cymbal?
dot = lower noise
exclamation = higer noise
space = pause?
they to my understanding do nothing, but act as unsung words
my issue is that I do not understand what people are trying to do in those examples
when someone writes [female vocal] its clear to me as a human what they want to do, but ai might do it (or not)
but when someone writhes !! ... ! ! ! I am to one who have a problem to understand what they want
i am trying to understand how to use those myself
I assume you cannot upload said clip because of the copyright content check?
It would be cool if there was a percussive meta language there...but I don't think there is
As Raymond and Alenas mentioned, these characters are patterns to the AI, not explicit instructions.
If you think about it: what would train the AI for making specific noises in regards to the unusual characters given? Almost zilch in the vast amount of resources needed for it to be trained with.
(I mean, you can do it, but I have my doubts it was done during the human supervised training phase)
And because AI is random in the thought process (especially since we cannot provide a seed for the generation), the assumption of what the specific character means can change upon generation.
So do not worry about what they mean or represent, as it can easily change, just simply use them as a pattern
i asked chatgpt to try to explain, and i like the answer
It looks like you're referencing a percussion break, which is a section in a piece of music where the percussion instruments are featured prominently. In written form, percussion breaks are often represented with a series of symbols or rhythmic notations that indicate the intensity and pattern of the beats.
Here's a breakdown of what this might represent:
Dots and Exclamation Marks: These symbols could represent different types of hits or accents. For example, a dot might indicate a softer or less accented beat, while an exclamation mark could denote a stronger or more emphasized hit.
Rhythmic Patterns: The arrangement of these symbols suggests a rhythm pattern. The spacing between symbols (e.g., between . and ! or !!) can indicate different note lengths or the timing between hits.
If we were to interpret this musically, it might mean:
The first line could be a sequence of softer or lighter beats (.) followed by a more accented beat (!), creating a rhythmic pattern.
The second line intensifies with more exclamation marks, which could indicate a crescendo or a more energetic part of the percussion break.
nice work chatgpt
Chat gpt does need ot be specifically told not to mention echoes, rising phoenixing, breaking chains, tapestry, whispers in the night, silent screams and cries, and above all NEONNNNNN
at the end of the day, you have to remember that chatgpt is just a language model that predicts words, its just analysed what youve given it and tried to predict what it is about - that doesn't mean suno understands it this way as Paimon said, its likely to be taken differently with each generation and could be influencing all kinds of things bcos suno doesnt rlly have specific pattern prompting like this. i wouldnt rely on using punctuation marks for anything specific.
To the point where you might as well write it yourself haha.;
the neon lights in ai chat, bugs me
I like to call AI in its current form "autocomplete on steroids", and believers in almighty ChatGPT hate me because of that
yes but i just uploaded to the offtopic chat if you want to check the differences
this is very clear for example
i would use 4 chars in all of them
5, 6 or wherever, just to be equal intervals
@naive quail this reminds me of some of the stuff youโve been getting up to
@viral moss Tokerโs been using styles like โaudiobookโ โspoken wordโ โinstructionalโ โstandup comedyโ to suggest to Suno that the dialogue is more important than the music, and itโs gotten some interesting results.
Of course, he usually adds the style prompt โGoats on Welfareโ, so that probably does a lot of the heavy lifting.
good fun!
Enraged Goats on Welfare! Get the band name right, foo!! 
omg, I just learned another piece of Toker-lore
Is there a way to do song to music style. If you want a new song in the same style
re-use prompt
change the title tho ๐
and lyrics if you have any, style of music will be that of the other song
I meant if I have an external song and would like how suno would define the style of the music
you'd have to define that, when you extended the uploaded clip
you can combine genres to get a new one
There MIGHT be some tools that can do that, but can you think of the way the song makes you feel? It doesn't have to be "correct terms", because you can "talk to Suno" to help it understand
For instance, if you wanted to make a song like Back in Black -- let's think of how we could describe that song -- "crunchy guitar", "off-beat rock and roll" "yelled" -- that's just at a glance, without even trying to try to describe a specific genre or a "feel", or anything more vaguely thematic
I know, but if I find a good song on youtube and want to write a completely new song but with the same style
Song analysis, lyrics analysis and music analysis. Understand what your favourite songs are about. Get accurate and in-depth music autotagging with SONOTELLER.AI, a tool powered by artificial intelligence including a comprehensive summary about the song lyrics and music. Analize your song, analyze genres, analyze moods and enhance your musical u...
Hey @crimson cave #๐โsuno-school message here's the link I was talking about in gen-chat
Thanks! I was scrolling like crazy!
Keep in mind this is just the armchair opinion of an AI-enthusiast -- not an expert on the matter, so feel free to tear anything that doesn't make sense or ask any follow-ups if you're curious โค๏ธ
I just find this kind of stuff fascinating, because if someone can think it these days, they're close enough to just being able to do it ๐
So i've done multiple remakes in the last couple days
using the new mys techniques
and tennis of course
an its basically a direct improvement
in both adherence IN song
as well as in the actual sound quality
the only ones which actually do not benefit is that reaaaally random prompt I have that rests on a knifes edge
using phonk
to produce that 1980s punkrock
post-punk
type
and I think that one is just so weird anything throws it off the edge
mary in the Mirror, Paper lanterns, both were able to be made in REAL hyper styles
whereas before they led to disastrous progressions
and SAD
massively improved sound
and then ohno just to see
#mys-ten-post-slowcore fieldbaroque-math
midwest mumble screamo
manic, introspective, cryptic, poetic
#MySpace swancore song. Listen and make your own with Suno.
post-post-slowcore fieldbaroque-math
midwest mumble screamo
manic, introspective, cryptic, poetic song. Listen and make your own with Suno.
like, I already have them far out there
so most of the benefit is seen in areas I couldn't touch before, namely hyper
Just finished reading and I reached similar conclusions, also, I think an Image Generation AI could be trained to make audio; I mean, you can convert audio into images as spectrograms; and the inverse process (spectogram to audio) is also possible. So if you train an AI like stable diffusion with tons of (finely tagged) spectrograms of songs, you could end with a model that creates spectrograms based in your text input, and then turn the spectograms into audio.
an easier way to explain if the person is familiar with math: its just function fitting.
consider yourself as a function mapping your inputs (your physical experience) to outputs (your actions).
Machine learning, in this context, is basically mapping these things together, such that if you were exposed to the stimuli again, you would respond "correctly" (more or less, its more complex).
Imo its one of the easiest ways to envision it
The way that makes the greatest sense to me just bc its an interest of mine is using natural selection. If you think of all the possible responses of a machine learning model to a given output, backprop can be seen as a method of winnowing away the implied pool of possible weights your model can have. It is in effect a form of selective pressure on a system (the system being represented by these weights) and as the model is exposed to this pressure, it converges in a particular direction.
And much like other selective systems, it runs into the same sorts of potential pitfalls, like over fitting (those birds with really useless wings that evolved due to maladaptive mating strategies) or local minima.
and a lot that I have internally, but I am unsure if there are technical terms for
so like, to be clear, AI does not cut up old works and remix them, it is much much much "deeper" than that
it would have to have an upper level understnanding, ie multi-modality, to do stuff like that. It is more like a savant.
a human savant, with no generalized comprehension, but with an extraordinary and gifted understanding of certain aspects of the activity is has been trained on that a regular person might miss simply bc we never assigned language to particular features
That's actually how the old open-source Riffusion worked.
It works like human brain, you don't need to have actual pictures of cows to recognize a cow when you see It, you simply learn the traits of a cow and It becomes a concept that allows you to recognize a cow when you see It.
I didn't know but, It makes sense
presumably, yes. I mean, I think its all picking hairs: we are function fitters ourselves, that much should be obvious, and our need for sleep imo is indicative that we will at some point realize that this is an intrinsic limitation of the shared aspect of ours+LLMs cognition ie the need to update weights
so I mean, ours is a lot different in HOW
namely, miniature chemical computers which have waaaay more complexity than the typical activation function
per neuron I mean
and possibly some sort of quantum shtufff
but irl, I'm a function
I must be, since I have input and output
thus, a machine learning model can simulate me (presuming we knew how to tokenize my senses lol)
thus, functionally, it is the same
Suno also does Censor Beeps, confirmed!
https://suno.com/song/3f90d566-6030-4941-92da-ca486fe6e374
Slow Rap song. Listen and make your own with Suno.
even if the way the bread is made is a bit differente
yall, I'm tired
I meant backprop
The other problem with AI is, artists have been showing their work everywhere in order to sell It, with the assumption that 99,999% of people that enjoy their art couldn't learn how to replicate It (for many reasons like time, skill, aptitude, etc) and the ones that can replicate would make slightly different artworks so no problem, but now we have AIs that can learn in hours what a human learns in years of practice, so the artists and/or copyright holders don't want their artworks to be used to train AI because, they know AI would 100% learn how to do It (while 99,999% humans wont) and that's not good for business.
So although AI and Humans can make original works based in what they experienced in a similar way; the learning capacity of AIs is a problem for the art business (I mean music, photography, paint, and video eventually)
They call It "copy" but It's not that different of what a person calls "inspiration"
Did you make this and if so... Do you have multiple free accounts?
naw man, that would be breaking TOS, I would never do that
Good good. Just checkin!
Wild that you brought that up, I remember a while back when Stable Diffusion was really taking off that someone had found ways of prompting spectrograms, and they were able to play them back -- I don't think the experiment ended up "having legs" as far as I know, but it was really neat to see what kind of img-2-img results it made, and seeing what similarities persisted between the results.
I wonder if anyone ever ended up taking that idea any further...
Furrybeans said open source project Riffusion worked that way.
ััั ะฐัะฐะบะฐั ะฟัะพะฑะปะตะผะฐ ััะพ ั ะฑะพะปััะต ะฝะต ะผะพะณั ะฟะพัะตะผััะพ ัะพะทะดะฐะฒะฐัั ััะตะบะธ
ั ัะตะผ ััะพ ะผะพะถะตั ะฑััั ัะฒัะทะฐะพะฝะพ?
ะฟะพะดะฟะธัะบั ะฝะฐะดะพ ะพัะพัะผะธัั?
ะธะทะทะฐ ัะพะณะพ ััะพ ั ััััะบะธะน?
ะญัะพ ะฐะฝะณะปะธะนัะบะธะน ัะฐั, ะธะทะฒะธะฝะธัะต (this is english chat, sorry)
ััั ะตััั ะฟะปะฐะณะธะฝ ะฟะตัะตะฒะพะดัะธะบะฐ
there is an attacking problem that I can no longer create tracks,
what can this be related to?
Do I need to subscribe?
Do you have credits left? https://www.suno.com/account
no
Refreshes at 0300 UTC+0 (<t:1713236400:t> in your local time)
Each day, free accounts receive 50 credits at this time ^
~6 hours, then more songs ๐ช
I was only able to create 3 songs of 3 minutes each.
#mys crunkcore, Slacker, field-punk revival, swancore #MySpace #tennis #2000s #birthday #lolz #rarxdxd song. Listen and make your own with Suno.
yall the opening riff
Or is it only on the first day?
this is some of the best work by the composer
one shot
i've seen
its too bad the clarity isn't tehre
You may need to wait and refresh the page to see if credits come back -- each "Create" = 10 credits, and gives 2 songs
So each day there should be 10 songs possible
Every day ^
Udio has a new AI model ๐
Oh cool! Always exciting to see what kind of advancements are coming out, everything is moving so quickly these days that you really need to appreciate different services to get a "bigger picture" view on what things are "flash in the pan" features, and what the real developments are. ๐
I think the mods try to keep conversation about non-Suno stuff to the #๐ฃ๏ธโchill-chat channel, but I'd be interested in chatting over there if there's differences you've noticed between the models?
I 've just generated 2 songs (free user), but they are a joke, miserable. I guess the model is not working well yet... ๐
well, competition is good for everybody, too bad Suno is SOOOOO lazy changing things that can be changed in hours regarding UI
Finding the right way to speak to the AI can be a difficult journey. May I suggest this website: https://rateyourmusic.com/ There you can search up artists or songs that you would like to create a similar style to, and they will have helpful little terms there that cover the style. I use it often to inform me of how to get closer to what I want.
Also the community here is pretty helpful, if you ask what you're trying to get and show what you're using to try to get it, they can help steer you closer to your goal.
Can anyone update me as to why the album covers / song photos kind of stink now?
IIRC, the artwork is influenced by both the style prompts and the content of your lyrics.
If you really want a specific artwork cover as a free member, I suppose you could try the old 3.0 trick but I do not know if that still works (which basically boils down to having the final line, after [End], describing what artwork you want in a 3.0 extend)
Did it improved because you shorten the lyrics or because you added baroque-math? How LLama understand you tags: What an intriguing set of tags! It appears to be a collection of genre-bending and emotive descriptors that might be used to characterize a piece of music, literature, or even a mood or atmosphere. Here's my attempt to break them down:
"Post-post-slowcore": This suggests a sound or style that builds upon the slowcore genre, which is characterized by slow tempos, introspective lyrics, and a focus on atmosphere. The "post-post" prefix implies a further evolution or subversion of the slowcore sound.
"Fieldbaroque-math": This is a fascinating combination of terms. "Fieldbaroque" might evoke a sense of ornate, intricate arrangements, reminiscent of Baroque music, but with a more experimental or avant-garde twist. The addition of "math" could imply a focus on complex rhythms, time signatures, or musical structures.
"Midwest": This geographic reference might suggest a connection to the American Midwest, which could influence the tone, themes, or aesthetic of the work. It might imply a sense of heartland simplicity, rustic charm, or a connection to the region's cultural heritage.
Very well described: "Machine learning, in this context, is basically mapping these things together, such that if you were exposed to the stimuli again, you would respond "correctly" (more or less, its more complex)."
But how did you introduced the druming pattern to Suno? Was it like that? [tact:8/4][drums:B!..B!..] LLama understand this like:
[tact:8/4]: This likely specifies the time signature and meter for the music. In this case, it's 8/4, which means there are 8 beats in a bar, and the quarter note gets the pulse. This is an uncommon time signature, often used in progressive or experimental music.
[drums:B!..B!..]: This seems to be a drum pattern or rhythm specification. The B!..B!.. notation might indicate a repeating pattern of bass drum hits, with the !.. representing rests or silence. The exact interpretation would depend on the specific drum machine or software being used.
Might be better to use this pattern: [drums:BSTTBSTT]: This appears to be a drum pattern or rhythm specification. The notation BSTTBSTT likely represents a sequence of drum hits, where:
B might represent a bass drum hit
S might represent a snare drum hit
T might represent a tom hit (or possibly a hi-hat tick)
But remember, we need to spice that with fills and licks.
I think you summed it up well enough.
https://suno.com/song/b68f8dd3-ad97-4906-a6b7-75702d3b3b79
metal, speedcore rap, power drums, aggressive, drums, drum, Drum and bass EDM, with sound effects, record screech song. Listen and make your own with Suno.
Yeah. I think the argument is pretty on target.
https://suno.com/song/b68f8dd3-ad97-4906-a6b7-75702d3b3b79
metal, speedcore rap, power drums, aggressive, drums, drum, Drum and bass EDM, with sound effects, record screech song. Listen and make your own with Suno.
darnit -- why do I always read the coolest stuff when I'm about to head off to bed?!
I don't know how I missed this, but I'd love to rap with you sometime about intermediate-to-advanced suno kinda stuff, I know you're sharing a lot of your independent work, but it would be neat to contribute my credits to some larger prompt experiment playlists if you had anything you wanted to try to figure out together.
Let's touch base at some point
oh man, this is actually really helpful -- still looking around the site, but I looked up "Positive Force" by Delicate Steve, and that's a great way to describe what I'd failed to adequately describe -- or at least it gives me ideas of what to pair with my "known suspects" re: style prompting
try for your self im not going to say it works or not
bar length : will slow suno gens alittle or some times longer
pick and chose things that will complement a genres
bye!
Please help! Suno ai won't generate me lyrics!
close and reopen your browser,
then report as bug to get your credits eventually back in the song area not discord
hello
i made an instruction book
copy-paste this into chatgpt, and it'll help you generate suno music
Are you uploading the file?
after copypasting this, do the following :
1/ type "generate dictionary for [artist/genre/vibe]" (warning : artists won't work too well since ais can only go so far)
2/ type "generate prompts", and chatgpt will give you 10 complex prompts you can use in the Genre of Music, in suno (alongside criterias to define the prompt, how original it is, how good, how powerful, etc)
3/ if wanted, type "make lyrics for prompt ...", it'll then generate its lyrics for the prompt and the lyrics should somewhat be relevant
OR, you can type "make metatag lyrics", and suno will attempt to generate exclusively lyrics based on metatags, to keep the song instrumental while giving it precisions
yeah if u want
i'll demonstrate one example with NCS, because i love ncs so y not
1/ type in the generate dictionary command
suno will then generate a dictionary of words for whatever you gave :
then 2/ type in the generate prompt command
(i'm only showing half)
chatgpt generates prompts, NCS-like, with criterias and potential words to add
if you also to output to scss format easy to copy into suno
fully configurable depending on what you want obviously, it's chat-gpt (if you want a specific instrument, more original, if you wanna exclude a theme, modify a prompt, it's all possible)
i then asked for lyrics, for the 2nd prompt, and chatgpt gave me this
after some quick copypasting work, here is the final result
just ask it to do it for you
wym
ask chatgpt to att brackets proper prompts and pbarackets
i did
for suno
i did
but chatgpt can only take so many instructions at a time and won't necessarily work for everything
for instance, it perfectly added the brackets on other generations I did
here's the part that handles this :
"[Name of the category]
High/Medium/Low relevanceย : word1, word2, โฆ
Do not skip a line between relevance and the associated words.
[Prompt 1 : word1, word2, โฆ / Originality = x1, Power = x2, Relevance to artist = x3 / Words that could be addedย : word1, word2, โฆ] Do not put any / between the words. Do not do genre / instruments / ..."
oops nvm
i forgot to specify the format for songs LMAO ur right
but so yeah if anyone wants this instruction book i manually typed suit urselves
what's that? i know nothing abt it
ask it to output yaml text then its clean less copy past
you guys are so smart, making your work easier LMAO
eh i was just bored
lmao
specify every thing
what type of stuff can I add to my notes? i'm really curious
well that makes sense but i don't rlly know how much I can specify with suno
since the ai modifies ur prompt each time anyways
copy paste the yami text into the lyrics?
ask chagpt what works what dosnt ask it abou ertain generas
how does it work?
in my format yes
how does it work can you explain to me
can you include tempo in suno?
well it pretty much sorts it out automatically for me based on the ton of words i put
so i didn't rlly think abt it
find out for your self , i wont say any more but iv played around with and i find it dramaticaly changes my song with the proper prompts
okay but you put this in the lyrics right?
yes
google all thos setting for the type of music you want it to feel like
i'm surprised that suno can actually understand this well
well like
im surprised it FOLLOWS it well
nice stuff
well
i have a section for generation
if i modify that, it should work fine right?
Section 3/ The generation
1/ When I ask you to, generate 10 prompts, while using each time the following format :
[Prompt 1 : word1, word2, โฆ / Originality = x1, Power = x2, Relevance to artist = x3 / Words that could be addedย : word1, word2, โฆ] Do not put any / between the words. Do not do genre / instruments / ...
2/ The first prompt must be a ยซย normalย ยป song. The others can have randomly chosen parameters.
3/ For each prompt, you can add a maximum of 5 words that could be added to further enhance the prompt, or slightly modify it.
that should be interesting to modify
Specifying specific keys and specific tempos will change the music, but you're probably not going to get those specific keys and tempos.
hmm i dont have a generation prompt let me see it that works
i'll give you my end results if you want
sure
well yeah suno modifies everything a bit
also hi mr moderator
but you're confirming it'll have an impact, right?
yes
i'm guessing suno slightly offsets everything to obtain something coherent in the end
Any word can have an impact and if you're using technical musical language that type of vocabulary as a whole probably shifts the music in a certain direction.
oh wow that's interesting
can i have some examples?
of technical language that would REALLY shift the music
bpm i'm guessing?
or mesures, or whatever
[note: generate: 5 versions]
[note: Song details and metadata follow]
[note: tempo: min: 60, max: 100 ]
[note: bar length: 6/8 ]
[note: tonic: Fmaj ]
...
also @tight spade i have a question
For example, if you use no style prompt and if you put no words in brackets, but if you write a very obvious sea shanty. Suno can still figure it out and sing the music as a sea shanty.
yes
Anyone else use the GPT created for Suno prompting?
i see that you wrote "people it should sound like" or wtv
but it sound like a donky singing
does that actually work
suno specifies the ai can't understand people
or smth like that
mhm mhm
I kind of understand the idea yes
It doesn't work in the sense that it doesn't make Suno sing more like those people. It does work in the sense that names can be understood as styles since you know, although which style it is is typically either very weak or very unpredictable.
okay so just unreliable
nice to know
But as far as words to use in a prompt names are pretty far down the list I think. Evocative regular words are probably more effective.
my guess is that, because suno and chatgpt are both AIs, there is something, like
context-lacking about them
which makes the communication and guess they give slightly more effective than what a regular human like me (who doesnt know that much about music) could do
which is why it actually works kind of efficiently
Ive had some pretty solid outcomes using a GPT
i don't know if that's right but yeah
It bassically enables me to really narrow down what I want
i can give u my instruction book if u want
Sure
here, if you wanna read
it basically enables you to make songs with some random commands
copypaste in chatgpt and ur good to go
idk if it's really needed but i made it and i'm happy i made that lmao
I mean, there is a pre made GPT that is already set up, you type what you want and it spits it all out.
suno helper
ask it questions
see what works what dosnt
and if you just want instrumental music don't put lyrics.
just your info for song info
puts out a nice instrumental
yo if i copypaste the details
[note: Song details and metadata follow]
[note: tempo: min: 128, max: 128]
[note: tuning: Standard (EADGBE)]
[note: genres: Future Bass]
[note: styles: Energetic, Uplifting, Powerful, Male vocals]
[note: instrumentation: Synthesizers, Drum Machines, Lead Synth, Electronic Drums]
those
into the song lyrics
i leave the style blank right?
yes
put title tho
no in box
ye all good
oh i got sum nice
more like this
it eats you cap for text but its worth it
then you liten to the song find where its about deviate from lyrics remember time then exstend from ther
im trying to get to stricly use a male voice but 1 out of two gens will be a duett
still nice
hum hum
lol
ima ask claude to just write it all for me rq
thanks claude
not done yet tho
now, structure all this information using the following format :
1/ parameters
for every single "note" parameter, write the following :
[note: (word): (word)]
2/ examples
for every single "note" parameter, write the following :
(word) : (every example associated to the word)
give me my magic
change all thos for your own unique songs and instrumentals or stuff that will suite your music
cool man
1/ Parameters:
[note: tempo: (tempo)]
[note: bar length: (bar length)]
[note: tonic: (tonic)]
[note: tuning: (tuning)]
[note: genres: (genres)]
[note: sounds-like: (sounds-like)]
[note: styles: (styles)]
[note: moods: (moods)]
[note: arrangements: (arrangements)]
[note: instrumentation: (instrumentation)]
[note: production: (production)]
[note: structure: (structure)]
[note: dynamics: (dynamics)]
[note: emotions: (emotions)]
[note: chords: (chords)]
[note: Vocals: (Vocals)]
[note: Vocal prompts: (Vocal prompts)]
2/ Examples:
tempo: min: 60, max: 100
bar length: 6/8
tonic: Fmaj
tuning: Eb Ab Db Gb Bb Eb, 440Hz
genres: Ballad, Pop, Soul, R&B, Contemporary Adult Pop
sounds-like: Male vocals, Sam Smith
styles: Male Singer, Emotional, Atmospheric, Soaring
moods: Hopeful Longing, Unwavering Resilience, Empowering Transformation, Joyful Celebration, Soaring Triumph
arrangements: Soaring, Storytelling, Cinematic, Textured, Clean Instrumental, Supportive
instrumentation: Cello, N Electric Guitar, Drums, Piano
production: Crisp T tones, Warmth, Rich Reverb, Complex Layering, High-Quality Recording, Dolby Surround
structure: Intro, Instrumental, Verse 1, Pre-Chorus, Chorus, Verse 2, Pre-Chorus, Chorus, Bridge, Chorus, Outro
dynamics: Smooth Transitions, Emotional Builds and Releases, Powerful Verses, Explosive Choruses
emotions: Deep Longing, Heartfelt Connection, Resilient Strength, Empowering Transformation, Joyful Celebration
chords: Amaj, Emaj, F#m, Dmaj, Bm7, E, Fm, Bb, Eb/G, Abmaj7, Gmaj, D/F#, Em, Cmaj
Vocals: Only male vocals. Vocals should be rich and emotive, ranging from soft and gentle to powerful and soaring.
Vocal prompts: soft and gentle, building intensity, powerful and soaring, gentle and reflective, uplifting, triumphant, reflective, fading out
copy pasting time lmao
idk how to use it but
eh
i can do my job still kinda
alr it works but
how do i write in this format you wrote in earlier
note: tempo: (tempo)]
[note: bar length: (bar length)]
[note: tonic: (tonic)]
[note: tuning: (tuning)]
[note: genres: (genres)]
[note: sounds-like: (sounds-like)]
[note: styles: (styles)]
[note: moods: (moods)]
[note: arrangements: (arrangements)]
[note: instrumentation: (instrumentation)]
[note: production: (production)]
[note: structure: (structure)]
[note: dynamics: (dynamics)]
[note: emotions: (emotions)]
[note: chords: (chords)]
[note: Vocals: (Vocals)]
[note: Vocal prompts: (Vocal prompts)]
thats the format ask it to imput what you wrote in to those fields
its smart it will do it
nice
only problem is
it fkin writes the lyrics in formatting as well
help
oh nvm i figured it out
tuning is very important
woodhouse is way too strict help
lol its ok
alr i incorporated them correctly
when i ask for hz, it spits out hz for every single note of a specific tuning
how do you come to the conclusion of having 440hz?
yeah i got it all
i also included potential lyrics for each generation
but it doesnt write em in markdown
example
then tell it to add [note : to meta or prompts at start
what is it for?

