Personally, I have little interest in typing a text prompt and getting a complete song as an output. However, I will gladly pay serious money for a tool that interactively collaborates with me in a granular, iterative process of generating, adjusting and mixing individual instruments and sections toward a finished multi-track song project.
Allow users to creatively engage by providing suggested starting places in the form of BPM, key and chord progressions or as brief audio and/or MIDI sketches. For example, let me give the AI a simple sketch of a couple bars of melody as MIDI notes, then have it give me back several variations of matching rhythm section and harmonies. Then take my textual feedback on adjustments I'd like but let me be specific when necessary, down to per-section or individual instrument. Ideally, the interface should look like a simplified multi-track DAW, making it easy for users to lock or unlock individual tracks so the AI knows what to keep and what to change as we creatively iterate. Once finished, provide output as both full mix and separate audio stems with optional MIDI for tracks with traditional instruments.
Targeting this use case accomplishes two crucial things. First, it lowers the bar of quality the AI has to match to be useful and compelling. Let's face it, generating lyrics, melodies, instrumental performances and production techniques more compelling than a top notch team of humans is hard for an AI. Doing it every time and only in the form of a complete, fully produced song is currently nearly impossible. The second thing it does is increase the tangible value the AI can contribute right now. Today it can be the rhythm section I lack, tomorrow it can fill in for the session guitarist I need, next week it can help me come up with new chord progression ideas. It would be useful every time I want to create music, whether I need backing vocals, a tight bass riff, scary viola tremelos or just some musical inspiration. And nothing it did would have to be perfect to be insanely useful - because I can tweak individual audio stems and MIDI tracks far faster than trying to explain a certain swing shuffle feel in text prompts.
Seriously, for a tool anything like what I've described, I'd be all-in for at least $30/mo if it's only half-assed. Once it's 3/4-assed put me down for $50/mo - and I'm not even a pro or semi-pro musician or producer, just a musical hobbyist who screws around making stuff no one else ever hears. Sure, actual music creators are a smaller market than music listeners but we're loyal, not as price sensitive and our needs for perfection in collaborators are far lower too. Plus, all those granular interactions as we iterate with your AI step-by-step towards "great", becomes invaluable training data - yet doesn't require us creators to surrender rights to our final output. For training data, the journey is (literally) the reward.
Honestly I don't think Suno knows who's supposed to be using their product. There's valid use cases for typing a text prompt and getting a song with zero fuss: background music in YouTube videos or corporate training videos, intro music for events (Starcraft fans know the ASL has been abusing Suno songs for a few seasons now), an AI-generated playlist running in the background in a cafe, etc. But those use cases don't seem worth too much, when you consider that music was effectively already free--you can hop onto Soundcloud and filter by Creative Commons and there's an unlimited number of songs you can use for these cases.[0] It remains to be seen how big the market for content creators who would actually pay for AI-generated songs rather than just pull CC tracks from Soundcloud is.
So then there's the casual end-user who's making music for themselves to listen to. IMO this is largely a novelty that hasn't worked out. I haven't heard many people regularly listen to Suno because, again, music is already incredibly cheap. Spotify is ~$15/month and it gives you access to the Beatles and Rolling Stones. The novelty of AI-generated "Korean goa psytrance 2-step" is fun for a bit, but how much will people pay for it, how many, and for how long?
I do think there's a lot of potential targeting musicians who incorporate AI-generated elements in their songs. (Disclaimer: I am a musician who has been using vocal synths for many years, and have started incorporating AI-generated samples into my workflows.) However as you point out, the functionality needed for Suno to work here is very different from the "write prompt, get fully complete song" use case.
It'll be interesting to see where it goes from here. In general, AI-based tooling does appear to be pivoting more towards "tools for creators" rather than "magic button that produces content", so I'm hopeful.
[0] One notable one is the artist "009 Sound System", who had a bunch of CC-licensed tracks that became popular due to YouTube's music swapping feature; since the list was sorted alphabetically, their tracks ended up getting used in a ton of videos and gaining popularity. https://en.wikipedia.org/wiki/Alexander_Perls#YouTube
> the functionality needed for Suno to work here is very different from the "write prompt, get fully complete song" use case.
Yeah, AI music gen is super fun to play with for a half-hour or so - and it's great when I need a novelty song made for a friend's wedding or special birthday - like, once a year maybe. But neither of those seems like a use case that leads to sustainable, high-value ARR. I'm starting to wonder if maybe most AI music generation companies ended up here because AI researchers saw a huge pile of awesome produced content that was already partially tagged and it looked like too perfect of a nail not to swing their LLM hammer at. And, until recently, VCs were throwing money at anything "AI" without needing to show product/market fit.
I'm not sure they fully thought through the use case of typical music listeners or considered the entrenched competition offering >95% of all music humans have ever recorded - for around ~$10/mo. As you said, another potential customer is media producers who need background tracks for social media videos but between the stock music industry offering millions of great tracks for pennies each and the "Fivver"-type producers who'll make a surprisingly good custom track in a day that you can own for $25 - I'm not seeing much ARR for AI music generators there either.
Currently the launch hypothesis of AI music generation puts them in direct competition against mature, high-quality alternatives that are already entrenched and cheap. And those use cases are currently being served by literally the best-of-the-best content humanity has ever created. Targeting replacing that as their first target seems as dumb as a SpaceX setting "Landing on Mars" as the goal of their first launch. There's no way to incrementally iterate toward that value proposition. Sure, targeting more modest incremental goals may be less exciting, but it also doesn't require perfection. Fortunately, music producers have needs that are more tractable but still valuable to solve - and not currently well served by cheap, plentiful, high-quality alternatives. And music producers are generally easier to reach, satisfy and retain long-term than 'music listeners' or 'music licensers'.
But I honestly don't see the point. The journey is the whole point when making music or any art really. AI doesn't solve a problem here. There never has been one in the first place. There is more music out there than you could ever listen to. Automating something humans deeply enjoy is misguided.
I actually used to have the exact same position when stable diffusion came out.
All my artists friends where criticizing it and I was thinking it was some form of Neo-Ludditism that they were following. Why not embrace progress? No one is stopping them to not use it but if it helps lower the barrier of entry isn't that great? Surely generative AI could be used to enhance of the workflow of an artist?
Oh, how I have been wrong. In reality it has only been used to replace artists. To devalue their work. It has not place in a artists pipeline.
I think the use of generative AI or at least of generalist LLM's is something fundamentally different than artists embracing new media and new processes. Like digital drawing is still roughly the same process as drawing on paper. The process is largely the same and most skills carry over. You are still in control. Using a prompt to create images is something that is not drawing.
I see the argument but can’t agree. To me it feels like pining for all of the lost typist jobs when word processors appeared.
Artists aren’t entitled to a job any more than I am; to the extent they (or I) do work that is easily replaced, by AI or offshoring or anything, that’s a career risk.
Those same tools enable artists to be more productive, more creative, more capable. The people suffering are those who are attached to the tools they learned with. We in tech know that’s a bad long term idea, and it is surprising in the arts. But on the whole, human creativity and overall output of great art will benefit from this change.
Your initial opinion was the right one. Too bad you let yourself be bullied into changing it by the vocal minority of online personalities. Artists that are not influencers have already embraced AI.
Generative tools and AI are great for finding inspiration. Same as using presets (which many do still think is "cheating" despite everyone doing it). Ultimately all listeners care about is the end result, not how it was made.
There wouldn't be a point in seeing artists play live if people only cared about the end result.
Listening to the studio mix on my headphones at home will always be better sound than being in a crowded concert.
I mean you are right to a certain degree, if it works, it works and if generative tools inspire you to make better music that is great. I am not so sure about that though.
I am forced to vibe code at work and it has not make me more creative. Is has made me want to quit my job.
Artists playing live is the end result though. If anything, the fact that people go to see Katy Perry in concert and not Max Martin, or how Tiesto is still a massive festival draw despite everyone knowing he's used ghost producers for 2 decades, are great examples of how little people care about the process of music being made versus the end result.
I'm not saying you need to use generative tools, but if it helps you make music you should do it. Ultimately what you're sharing with the world is your taste, not your technical abilities. To slightly expand on a famous quote in the music world -
> I thought using AI was cheating, so I used loops. I then thought using loops was cheating, so I programmed my own using samples. I then thought using samples was cheating, so I recorded real drums. I then thought that programming it was cheating, so I learned to play drums for real. I then thought using bought drums was cheating, so I learned to make my own. I then thought using premade skins was cheating, so I killed a goat and skinned it. I then thought that that was cheating too, so I grew my own goat from a baby goat. I also think that is cheating, but I’m not sure where to go from here. I haven’t made any music lately, what with the goat farming and all.
i'm not really into music, at all, but would easily pay $50/month to be able to have this functionality, too. even just the tools being what they are i've spent over a dozen hours playing around with it because it's so much fun. add the ability to separate parts of the tracks and select portions specifically to change and i'd use this constantly, all the time
Thanks for the suggestion. It looks closer to what I want than anything I've seen so far. Probably still not granular enough but much closer and headed in the right direction.
Allow users to creatively engage by providing suggested starting places in the form of BPM, key and chord progressions or as brief audio and/or MIDI sketches. For example, let me give the AI a simple sketch of a couple bars of melody as MIDI notes, then have it give me back several variations of matching rhythm section and harmonies. Then take my textual feedback on adjustments I'd like but let me be specific when necessary, down to per-section or individual instrument. Ideally, the interface should look like a simplified multi-track DAW, making it easy for users to lock or unlock individual tracks so the AI knows what to keep and what to change as we creatively iterate. Once finished, provide output as both full mix and separate audio stems with optional MIDI for tracks with traditional instruments.
Targeting this use case accomplishes two crucial things. First, it lowers the bar of quality the AI has to match to be useful and compelling. Let's face it, generating lyrics, melodies, instrumental performances and production techniques more compelling than a top notch team of humans is hard for an AI. Doing it every time and only in the form of a complete, fully produced song is currently nearly impossible. The second thing it does is increase the tangible value the AI can contribute right now. Today it can be the rhythm section I lack, tomorrow it can fill in for the session guitarist I need, next week it can help me come up with new chord progression ideas. It would be useful every time I want to create music, whether I need backing vocals, a tight bass riff, scary viola tremelos or just some musical inspiration. And nothing it did would have to be perfect to be insanely useful - because I can tweak individual audio stems and MIDI tracks far faster than trying to explain a certain swing shuffle feel in text prompts.
Seriously, for a tool anything like what I've described, I'd be all-in for at least $30/mo if it's only half-assed. Once it's 3/4-assed put me down for $50/mo - and I'm not even a pro or semi-pro musician or producer, just a musical hobbyist who screws around making stuff no one else ever hears. Sure, actual music creators are a smaller market than music listeners but we're loyal, not as price sensitive and our needs for perfection in collaborators are far lower too. Plus, all those granular interactions as we iterate with your AI step-by-step towards "great", becomes invaluable training data - yet doesn't require us creators to surrender rights to our final output. For training data, the journey is (literally) the reward.