Adobe announced new software for music creation using AI.
Adobe's Project Music GenAI Control can generate music from prompts and allows editing elements like tempo, structure, intensity.
Just like popular AI apps like Chat GPT that are based around prompts, Adobe does the same.
Enter in a prompt like "lofi hiphop" or "upbeat rock" or "sad jazz" and it'll churn out a sample for you.
According to Nicholas Bryan, senior research scientists at Adobe:
“With Project Music GenAI Control, generative AI becomes your co-creator. It helps people craft music for their projects, whether they’re broadcasters, or podcasters, or anyone else who needs audio that’s just the right mood, tone, and length
There's a few tools out there that do this. The kicker for Adobe's Project Music Gen AI is it goes beyond simple generation by allowing users to refine and shape the results.
With this tool they can adjust the tempo, increase or decrease the intensity, create a seamless loop, create related song structures, etc.
Great For Small and Independent Musicians
Tools like this can be good for small and indepdent musicians. It can serve as a time and money saver compared to searching for existing samples. The tool could be used as an inspiration generator to help spark new ideas or get out of creative ruts.
It also helps to avoid copyright issues
AI and Copyright
Fully AI-generated works are currently considered non-copyrightable, so using samples from the tool avoids potential legal issues around copyright infringement. This one of the benefits for small and independent producers using tools like Adobe's Project Music Gen AI.
The ability to quickly get a sample that you know is not considered copyrightable will help producers speed up their workflow.
Moral Issues With Music Generated by AI
There are debates around the moral issues of AI-generated content being built on the work of human musicians without consent. This obviously leads to some musicians feeling not so great about it.
There are also the huge issues of people using AI to completely mimic a famous artist's voice. We've seen this play out with Drake, The Weekend, and Ariana Grande. Celebrity artists like this obviously are not cool with it.
Adobe claims to only use public domain content to train the AI tool. This is nice to see Adobe using AI in a responsible manner.
How's the Quality?
So far the sounds I've heard in the demo are pretty basic. Not bad, not great. They definitely sound like backing tracks, or generic tracks. Again, not bad, but they won't be hit makers.
But the reason for that - I would guess - is because the model is trained on public domain content. IE, it's using music that is royalty free. Or in other words, Adobe isn't ingesting Timbaland beats to model it's software.
I would imagine at some point, Adobe would bring in top producers to help train the algorithm. Maybe not Timbaland level, but top producers who may not be so known, but are producing for A list acts.
At that point I would imagine the output would be much better.
Fading Out
Adobe's Project Music Gen AI gives insight into how useful AI could be as an endless source of inspiration free of copyright concerns.
No details as of yet when it'll be released but we'll be keeping an eye on it. Also looking forward to the real name - Adobe Project Music GenAI is....a lot lol!