Use the 'continue from this clip' feature. Next, merge the newly generated segment with the existing one using the 'get whole song' feature.
Or just get creative, use the vocal remover to extend the background sound, cut and paste different parts of it using the good old DAW. By using this method, i managed to make several "full" songs :
Our models support full 2+ mins of coherent generation but generating a couple of verses at a time through continue gives good results you can keep picking the continuations that sound best!
Using Suno(custom) and ChatGPT [0] + editing - I just helped heal a lost friendship. I had no way to break the ice of our lost friensdhip for two years now. But this did it.
I was finally able to communicate what happened and I finally got a response from him.
I cannot describe how important this was for me. Thank you. I already subbed, but I would pay a lot of money to get the remainder of the song. Don't make me open Ableton.
Whoa! Did you guys just change the voicing that is singing in the link? Wait, did the track change entirely? Does this happen every so often?
I don't get it. Why would it regen? Should I be saving audio and sharing it via GDrive or whatever at every interesting playback?
I was replaying this track all night long. Then suddenly the linked song was different. This is not good UX. I shared it with my friend, which version will they hear?
Each time I hit Create, it should make a unique waveform. Crappy from an ML pov, or not. This is all a matter of listener preference. Sometimes artifacts are cool.
Each shared link should be a unique "printed" MP3 distributed via CDN, right? Or, am I being silly, and why? This also sounds so much cheaper as far as server costs, doesn't it?
Last of all, I love this entire thing. If you have any room in hiring, I have spectral opinions and would love to be involved.
If you scroll through your library (https://app.suno.ai/me/) you can probably see multiple versions if you did make them ... eventually we'll have search :)
I see now in dev tools that it is indeed an mpeg being downloaded. Maybe it may have been a bitrate issue with my external sound device which may have confused me. Thank you.
This is the entire future of mass entertainment. A non-stop stream of endless content generated from the long tail of your digital advertising footprint. No more scrolling Netflix for hours, or waiting two years for a new season. Just an unending feed of everything you ever wanted to see or hear forever.
It will allow more people to do a Monte Carlo search across a much broader spectrum.
The mass will filter through this by popularity.
And good people/artists will be able to learn faster, more and iterate faster over ideas.
For everyone else who actually doesn't care that much, they will get similar content cheaper.
After all there are so so many people watching normal TV daily with a ton of advertising or blindly radio which delivers the same top 50 list over and over.
A little feedback i hope the Suno team sees. Took me a long time to figure out how to play the songs. Hitting play at the bottom didnt work. Eventually I figured out to mouseover and hit the arrow play.
Another piece of feedback which might save you server costs:
On /create/ in custom mode, after tapping the Create button, I feel like I should be shown the progress in the Library to see the result processing.
I am surely more likely to want to see the output after Create is tapped than to stay at that screen to create another immediately without see the previous result.
When not in custom mode, the user sees the processing immediately. In custom, I will keep tapping Create for no good reason thinking that nothing happened. [0]
Also, this is super fun to the point where I actually subscribed. Thanks for making this!
[0] Currently the button changes label for 1 second, but that's not what I want to see. Before figuring it out, I wasted your my credits and your server costs by tapping Create unnecessarily.
>On /create/ in custom mode, after tapping the Create button, I feel like I should be shown the progress in the Library to see the result processing.
You don't see a progress bar but you should see the song appear with a loading icon in about 1 second. Might be a UI bug. What are you viewing the site on?
Please email my username at the big G, I would love to help with this magic. I can't believe it took this long.
I have been fortunate enough to be near Funkadelic players. My point is that I am a little shocked that this took so long. Music is not that complicated.
I don't work for Suno, I just Chirp and Bark a lot. You'll have to contact Suno via email, or reply to one of the Suno team that popped up in the comments here.
Thanks for responding, and also thanks for making a truly useful product for people who communicate via music. You made a significant difference in my life.
There is quite a potential for ephemeral musical messages and memes. It can generate resonable lyrics similarly to ChatGPT generating responses with just “Give me funny response to ´claim´”.
In custom generation at least, when you tap Create, the button does not change or disable. So it's easy to use up credits on the same thing thinking that nothing ever happened.
Meanwhile, they all do get created on each tap and appear in your library.
There are some click artifacts, but this came out pretty well:
Let’s look at the input required for each, in the context of computer-aided music production.
‘Generated’ music: typing a prompt and pressing the RETURN key (time required: ~10 seconds)
‘Made’ music: thinking about melody, harmony and rhythm and writing down or programming in each from scratch. Choosing sounds and MIDI instruments. Experimenting with different effects, tweaking every parameter precisely and playing around (using your ears for feedback) until you find something you like. Finally, mixing the tracks together so the whole thing sounds cohesive. (time required: ~1 hour, min.)
To me, it seems hard to confuse the two processes.
You may very well argue that the ‘AI’ is doing something analogous to human music production (well… architecturally it isn’t, but you could at least argue it’s equivalent in some sense), but arguing that the human who typed the prompt ‘wrote the song’ seems to be… to put it lightly, rather overstating it.
You don't have much control over what is being generated. And if you did want to have fine control then what would be the difference compared to just dicking around on a synthesizer?
A lot of the enjoyment of music comes from connecting emotionally with the artist. The artist had something to say based on their experience of life and adversity. You relate maybe because you've been there too. After all, you and the artist are both just human at the end of the day.
This is why for instance I have a hard time believing in AI girlfriends or AI therapists. It's not that I don't think that an AI could learn to be empathetic and say the right things at the right time, it's that I think there is something about you the human knowing it is an AI speaking that would make you not be able to connect. It's the knowledge that they haven't had any life experiences like you. They haven't had adversity or struggle.
> You don't have much control over what is being generated
I don't understand this nitpick at all. What part of "make" implies fine-grained control over the output? Parents didn't have much control over what's generated when a child is, but by any reasonable definition of the word "make", the baby that comes out is "made" by the mother and (to a lesser extent) the father!
> A lot of the enjoyment of music comes from connecting emotionally with the artist. The artist had something to say based on their experience of life and adversity. You relate maybe because you've been there too.
I think you might be assuming that your experience generalizes a lot more than it does. I've been a musician for decades, and I'm constantly listening to music, and I don't need to know who did what to be able to enjoy a song. To be clear, there's nothing _wrong_ with that being part of your enjoyment, but there's nothing wrong with just liking to hear sounds that sound "good" without caring about where they came from.
I am definitely generalizing. It does seem to be a common sentiment expressed when this gets brought up though. Whether it's just an instance of me clinging on to anything to distinguish from AI or if it's a legitimate problem remains to be seen I guess.
And there definitely are songs I feel I just enjoy in their own right.
As for 'make' I guess that just comes down to semantics.
How do you reconcile the fact that music from people who have done pretty despicable things can still be widely popular? You can certainly argue that enough people are willing to look past someone being a murderer or child abuser or something because they still can find some sort of human connection with the artist, but given how uncommon that sort of sentiment is in basically every other part of life, it seems far more likely that people just don't really think about it an artist at all when they listen to their music if the artist is someone they wouldn't want to connect with.
I don't think his personality was really that much of an outlier when his music first became popular, so I'd argue that empirically the answer is yes. From looking at Wikipedia, his first album was released in 2004 and charted at number 2, with his first single hitting 15 on the charts and his follow-up being number 1. I don't think you can argue that he was famous enough for anyone to be listening to the music because they knew who he was rather than because of the music itself.
Nobody needs to know who the artist ahead of time is to connect to what they say (instrumentally or vocally). The art does the talking.
I largely agree that if you don't know anything about music but generate some stuff with a very-high-level AI tool then you are unlikely to produce anything that resonates with people for any significant amount of time.
If you do know something about music (say, producer- or other-tastemaker level) and you replace the artists with an AI tool - you could have much better luck - but I'm curious there how much longetivity you get. Could you create the next star or the next trend or will the tools not have the ability to "break the mold" in ways that really connect to audiences and new generations without being used by newcomers themselves?
> I largely agree that if you don't know anything about music but generate some stuff with a very-high-level AI tool then you are unlikely to produce anything that resonates with people for any significant amount of time.
I feel like you're not really making a strong assertion here because of how subjective "resonates with people for any significant amount of time" is. Instead, I'd propose something akin to the Turing test; instead of conversing with someone and trying to determine if they're a computer or a human, the participant would listen to to a piece of music and try to guess whether it was made "traditionally" or by someone who used an AI tool and had no experience making music in any other fashion. I think we're not far from the point where it would be possible to generate instrumental music with AI that would be indistinguishable from a control set of human-created music (either instrumental or with the vocal track removed from the mix) with a certain level of complexity (let's say songs without changes in tempo, time signature, or key, which would give us at absolute minimum a few thousand popular mainstream songs over the past half century, and potentially a lot more). How long do you think it will take for this to be possible (if ever)? If you don't think it will ever be possible, why not? And if you do think it will be possible, isn't this sufficient evidence that there isn't any inherent need for a "human" element in music?
It’s not about that. It’s more if ye uses ai to make music it’s still ye making the music.
Ye is actually an endorsement of this because he’s absolutely a creative director more than a skill based musician. His best works are from leading others to greatness and building situations for that rather than skill in strumming a guitar or whatever
> A lot of the enjoyment of music comes from connecting emotionally with the artist. The artist had something to say based on their experience of life and adversity. You relate maybe because you've been there too. After all, you and the artist are both just human at the end of the day.
Absolutely not the case for me. The artist is just a name and I often don't know what the song is called. I have zero interest in the "meaning". It's all about the melody/harmonics, beat and production value for me.
Melody, beat, and production techniques are all ways to express meaning. Do you feel anything when listening to favorite music? Odds are that feeling, or something like it, are intentionally conveyed, even in instrumentals.
Check out aiva.ai if you want more control over what’s being generated.
Regarding the connection with an artist - I think it’s overrated. I don’t really care about Lady Gaga life experience to enjoy her songs. I have no idea who created half of the songs on my Spotify playlist. Artists create brief virtual life experiences through their songs. Songs I like usually remind me about something I have experienced or would like to experience.
I agree about music. When it comes to girlfriends and therapists when the medium is text the connection could be faked with a large corpus and RLHF. Keep in mind that some people already have inflatable girlfriends. It aint pretty but may serve some purpose for disfunctional people. I’d get more worried if it becomes a bigger thing.
When it comes to chat therapy, it could be an interesting mode of self discovery. My only worry is if the goal is to attempt to replace therapists altogether.
Don't you think simply knowing it was an AI would mean you can't take it seriously? Or do you think this is just a cultural thing that will fade away in coming years.
Oh I hate those pseudo-smart replies like this. You clearly know when it's generating and the distinction between creating and making (composing) music. It's generating when you prompt the ai to do all the work for you. As simple as this, you don't touch the actual producing part yourself. If you commission a song to a producer to make for your game, you don't make the music. The producer makes it
Because humans aren't actually doing the generating either, the answer gets clearer when the question is something like, "are you making music, or telling something to make you music?"
Replicators are a good analogy. Ordering a meal from a replicator doesn't make you a cook any more than giving Midjourney an order makes you an artist.
If the number of different knobs is is sufficient to make it impossible that any one person could try all the combinations, manipulating the knobs could be an art. You could imagine eg cooking using that method.
And when we look at great songwriters, do we need to know their educational background and what music they’ve listened to in order to determine how much of their work is created out of thin air versus arrived at by reasoning over theory and inspiration from other work?
I have been generating songs for days now ; it's really good at some things, especially metal and rap. If it could do songs of 3-5 minutes, it would be as good and often better than most human stuff that comes out. And it can do any language; in Dutch it often makes the singing Flemish or Brabant/Limburg for some reason. The Spanish rap is excellent and got some great results with industrial metal too.
This is, at least for me, far more impressive than image generation. Once they have an API , i would just leave this running in the background to generate music while working.
This is nuts! If this advances as fast as AI image generators have, the music industry is going to be affected. This is a crazy tool for song writing inspiration for sure.
It's not at all clear what this actually is. The "About" section says basically nothing. It would be nice to know that before having to create an account.
What's the legal status of ToS clauses about restrictions on what you can do with the generated audio? I'd have thought that AI can't own copyright -- what does that say about one's ability to monetize (under the free tier, say) -- are those prohibitions enforceable?
(I can make plenty of guesses myself, so I'm most interested in hearing informed replies with references rather than speculation.)
This is a contract violation, not a copyright issue. But if you share your audio with someone who isn’t party to this agreement, there is likely nothing stopping THEM from using it commercially.
I think the low 'hit' rate might be turning off some people, but I'm happy to audition through 10 songs to find a single good one.
I hadn't been following the state of audio generation with AI, but this definitely feels like a breakthrough moment for me, just as image generation did, and chatgpt.
I'm a video-game developer, so can imagine using this now and then for quick jam games.
A fun way to use this I discovered by accident: ask ChatGPT to write a song for you. (You can use a longer prompt, and get a text version of the verses and chorus.( Paste that into the “Custom Lyrics” UI to get a musical version.
Would really like something like this to generate infinite lofi to code to. I'm struggling to get this to not output vocals though - I've tried both instrumental and no vocals in the prompt.
While it can only do american popular music styles its already useful for making samples that are at least not samples of straight up copywrited material
also NAL I'm not sure that holds... there's no "anti-copyright". I can take the lyrics, the melody, and register something. How is suno.ai say I can't copyright it if they don't own it themselves?
We won't know for sure until there's actual law or court decisions to reference, but until then, it's worth looking at something like the Monkey Selfie copyright dispute[1] for guidance. The article references a 2014 opinion from the United States Copyright Office
> only works created by a human can be copyrighted under United States law, which excludes photographs and artwork created by animals or by machines without human intervention ... Because copyright law is limited to 'original intellectual conceptions of the author', the [copyright] office will refuse to register a claim if it determines that a human being did not create the work. The Office will not register works produced by nature, animals, or plants
The extension of this to AI would be saying basically that the copyright office simply wouldn't extend copyright of an AI created work to any party.
In the case of one of these songs, though, if you wrote your own lyrics, you would still have the copyright to those lyrics, if not the full piece of music generated from them.
I think this is intended for advertizers. I have seen lot worse music mostly for products aimed at kids and they likely take 1000s of dollars to write and record.
You aren't comparing the same thing. Custom produced thing is entirely different than buying samples. You still need musical talent to work with library music.
I agree that the quality isn't ideal, but I think this tool helps artists iterate much faster and cheaper. I wouldn't focus on the quality of the output, beyond the threshold which allows the artists to generate a reasonable idea of what they eventually want to make.
Think about all the hard work that traditionally goes into composing a single title. Artists will spend days, weeks and sometimes months trying to iterate on ideas. Writing, composing, demoing, tracking and recording, mixing, etc. Think about all of the expensive software and hardware that goes into this process (instruments, microphones, studios, DAWs, VSTs, etc). It's an expensive and difficult process, it's very manual, very sequential.
This could easily be used to speed up that iterative process. Just ask this software to generate 100 ideas for your next bridge, and iterate that way.
I find this very useful as someone who's just learning how to play the guitar. My knowledge of music theory is still limited, and it'll take me years to get to a place where I can express myself in a way I'd deem "satisfactory". I just visited this page and plugged in my lyrics, and it arranged them into a beautiful song for me. It did it just how I imagined it, and that's terrific. Now I can ask my guitar teacher if the chord progressions make any sense, and if so, then we can transcribe it. I don't know who else this would be useful for, but I could see myself paying for it depending on how they develop the tool.
AI music generation sucks and I expect it to suck for a long time. It's hard to make good music. Even a fake band from a mocumentary show is infinitly better (Bill Hader, Fred Armisen, Documentary Now)
I don’t think your analogy to the band from Documentary Now (clearly music made by a very much real band, albeit satire) holds up as well as you think it does.
I checked with my current favorite AI (Pi) and you are correct. I stand corrected but still stand by my point. I even prefer The Shaggs over the AI music I have heard. It has personality.
Are synth sounds that emulate strings dehumanizing because they don't take years to master? Is the electric guitar fake? Is the arrangement of a six string guitar's strings in standard tuning 'yuck' because it impresses a certain chord voicing, taking away compositional decisions from the musician?
Then you got to explain to me how you define "instrument"? Because what you're describing is kind of the equivalent to comparing a cake to an "ingredient".
Ja, the tracks are quite average - but like I just commented in the thread above - maybe one day using art from AI would be like shopping from Amazon (instead of the local store down the road). Crazy times to say the least - impressive tech though.
It's just another tool - synths are computer generated as well, so what does it matter whether a computer or human decides to push the button?
At the end of the day it's humans that will use the tool and combine it how they see fit and it's humans that will decide whether it's any good or not ...
I totally agree with you actually - but the tooling here is for producing content, not for composing music (like your synth example). It's exactly the same as MidJourney - the target user is not the artist, but the regular user. I wouldn't be surprised in a couple of years we have terms similar to "buy local", but for content generated by AI - like "buy human".
https://app.suno.ai/song/e5c4214b-5851-4ae8-876b-6a99f2ca878... https://app.suno.ai/song/0425b83a-2fa7-4944-9ed8-ba554bd42c3...