An independent artist’s recordings are no longer just something to release, stream, and pitch for sync. They can also become part of the material used to train generative AI music tools, and that can come with a fee and a share of revenue, or with questions about consent, control, and what happens after a recording has been used.
That choice came into sharper focus on November 18, when AI audio company Sonilo announced a licensing deal with independent distributor Symphonic. The agreement allows Sonilo to train its AI music model on recordings from Symphonic’s independent artists and labels, but only when rights holders choose to opt in.
That last part matters. This is not a blanket license covering Symphonic’s entire catalogue. But an opt-in is still a business decision, and it’s worth understanding what you’re agreeing to before making it.
What the Sonilo–Symphonic deal does, and doesn’t tell us
Sonilo generates music and sound effects from video footage, text, or both. It pitches its platform to short-form and short-drama creators, developer platforms, and enterprise production teams. Its distribution partners include fal, TapNow, and ComfyUI.
The company was founded by former TikTok staff. Co-founder and CEO Shawn Song was previously a multimodal AI tech lead at TikTok, while co-founder and COO Keli Li built and led global music at TikTok. Based in San Francisco, Sonilo raised USD $11 million in October 2026 in a round led by B Capital, with participation from Redpoint.
Sonilo and Symphonic say participating rights holders will receive licensing fees and a share of revenue. Sonilo also says it will provide attribution reporting to support transparency and compensation. The attribution component builds on Sonilo’s partnership with Musical AI, announced October 27. Musical AI’s technology examines each piece of audio Sonilo generates and measures the contribution of licensed recordings and compositions behind it.
That’s a meaningful framework to discuss: licensed material, payment, and a system intended to track contributions. But it is not the same as having all the contract terms in hand. The financial terms of the Symphonic deal have not been disclosed, and the companies have not said how many Symphonic artists or labels have opted in.
Sonilo says it trains only on licensed catalogues and names Shutterstock as an early licensing partner. It has also said it is finalising master recording and publishing deals with major rights organisations, expected to close in the coming weeks. No such agreement has been announced yet.
In other words, the direction is clear, but important details remain open. That’s a good reason to read carefully, not a reason to assume the offer is automatically right or wrong for every artist.
Start with the rights: what exactly are you licensing?
“Your music” can mean more than one thing in a rights agreement. A master recording is the particular recorded performance. A composition is the underlying musical work, including its songwriting and publishing rights. They may have different owners, co-owners, or representatives.
Before you opt in, ask whether the permission covers a recording, a composition, or both. If you own or control only one set of rights, make sure the agreement reflects that. If there are co-writers, featured performers, a label, or other rights holders involved, check whether their approval is needed too.
This distinction is especially important because attribution can involve both sides. Musical AI’s technology is intended to measure the contribution of licensed recordings and compositions to generated audio. But an attribution system can only support payment for the rights and people it can identify. Clear ownership records and accurate registrations help make that possible.
Ask how far your permission reaches
An agreement might apply to a particular model, a particular catalogue, or a defined use. Don’t assume those limits, or assume there aren’t any. Ask whether the license is for one model now or could extend to future models, products, or uses.
Also ask what “opt in” means in practice. Which recordings are included? Can you select individual tracks, or does the choice apply more broadly? Who is authorized to give permission for the material you control? These are contract questions, not details to fill in from a headline.
Keli Li has described the vision this way: “the future of video-to-music is soundtracks made for every story, and that future has to be licensed from the ground up; independent artists deserve a seat in the AI music value chain.” Symphonic CEO and co-founder Jorge Brea has said AI is only a real opportunity for artists when it is built on consent, transparency, and fair payment, and that the deal gives independent artists and labels a direct, paid way to take part.
Those are strong principles. The agreement you consider should spell out how they work for your music.
Find out how and when payment works
The companies say rights holders get licensing fees and a share of revenue. The public announcement does not disclose the rates or payment schedule. That means you should look for the specifics in the terms you are asked to accept.
Ask what the payment is based on, how attribution affects it, and when money is paid. Ask what reporting you’ll receive and how you can review it. A promise of attribution reporting is useful, but it doesn’t tell you on its own how a fee is calculated, how often the report arrives, or how a dispute is handled.
Understand what leaving means
If you change your mind later, can you withdraw? If you can, how does that work, and what happens to models already trained using the licensed recordings?
Those are reasonable questions to ask before agreeing, particularly because removing a track from future licensing may not necessarily reverse what has already happened. Don’t rely on assumptions about takedowns, deletion, or retraining. Look for clear language about withdrawal and the treatment of material already used.
It’s also worth checking whether the permission conflicts with anything else you’ve agreed to. Review your distribution deal, publishing arrangements, and any agreement with a sync agent or other representative. A right you’ve already granted elsewhere may affect what you can license now.
Make your catalogue easier to identify, and keep your own record
There’s practical preparation you can do before any opt-in. Make sure your split sheets are complete and your works are registered accurately. Confirm that the names and ownership shares attached to your recordings and compositions are current. If a track has collaborators, get the paperwork in order rather than relying on everyone to remember the arrangement later.
Then keep a written record of every AI licensing opt-in you sign. Save the agreement, note which recordings or compositions it covers, and keep any confirmation or reporting you receive together in one place. A tidy record makes it easier to check what you agreed to when another opportunity comes along.
Symphonic’s deal with Sonilo is part of a wider set of opt-in AI licensing activity at the distributor. Symphonic has also announced an opt-in programme with attribution company Sureel AI, a SourceAudio deal, and MusicShield, a tool designed to make recordings harder for generative AI models to learn from without permission. Those options point to a changing landscape, not a single decision that settles every question about your catalogue.
The same technology can cut both ways
For an independent artist, AI training can become a paid revenue line, or an unpaid training input. The difference may depend on rights, consent, and the terms attached to a form that’s easy to click past.
That doesn’t mean you need to reject every licensing opportunity, or accept one just because it promises a new income stream. It means you deserve to know what you’re licensing, how payment is supposed to work, and what control remains afterward.
Start by checking what you may have already opted into across your distributor, publisher, and sync agent. Decisions can stack, and the one you remember making may not be the only one shaping how your music is used. From there, the best next step is a careful review of your agreements and a clear record of every yes.

