Dated voiceprint record
Register your reference recordings and stems. Each one writes a content hash to Base tied to your account at a specific block height: proof of what your voice sounds like and who claimed it, dated before any dispute.
Creator Protection
AI voice protection for musicians and vocalists.
Modern voice models can build a convincing clone of a singer from under a minute of audio, and every released track is source material. Suede gives your voice a dated ownership record and consent terms that software can read, so authorized use is licensable and unauthorized use is checkable against a public register.
Every track you release is a training set.
Voice cloning stopped being a research demo years ago. AI covers of well-known vocalists circulate on every major platform, and the tools that make them are free, fast, and improving. A working vocalist’s entire catalog, plus every interview and live recording, is available to anyone who wants to train a model of their voice. None of it required consent, and the resulting clone can sing words the artist never said, in contexts the artist never approved.
The law is catching up unevenly. Your recordings are protected by copyright, but the sound of your voice is a right-of-publicity question governed by state law in the United States. Tennessee’s ELVIS Act made unauthorized AI voice simulation explicitly actionable in 2024, and other jurisdictions are moving the same direction. Whatever the venue, the practical fight comes down to evidence: which recordings are yours, when you claimed them, and what use you authorized.
Vocalists face the question from both sides. Some want AI use of their voice shut down entirely; others want to license a voice model on their own terms and get paid for it. Both positions need the same infrastructure: a dated ownership record and terms that the systems using voices can actually read.
What a registered voice gives you.
Machine-readable consent
Your registration carries terms software can read: whether training on your recordings is permitted, whether synthetic vocal reproduction is permitted, and what an authorized use costs.
Derivative tracking
An authorized voice model or AI cover registers as a child of your reference works. The lineage is public, so a licensed clone is distinguishable from a bootleg by anyone who checks the chain.
Royalty routing
If you license your voice, the split executes on-chain. You, your co-writers, and your producer receive the shares specified at registration, without a platform accounting department in the middle.
Refuse, or license on your terms. Both need a record.
A public statement that you do not consent to voice cloning is written for humans, and training pipelines do not read it. The systems consuming your catalog are software, so your consent has to be software too. When you register recordings on Suede, each asset carries machine-readable license terms: training permitted or not, synthetic vocal reproduction permitted or not, and the price of an authorized use.
If your answer is no, the register makes your refusal legible: dated, public, and attached to the exact works in question. If your answer is yes on your terms, the same record becomes a storefront. A platform building licensed voice features, or an AI agent sourcing an authorized vocal, can query your ERC-8004 asset on Base, read the terms, pay the contract, and the royalty split executes to everyone specified at registration.
Either way, the alternative is silence, and silence is what unauthorized users count on. A voice with no record and no terms reads as free raw material to every system that encounters it.
Frequently asked questions.
- Can Suede detect or block AI clones of my voice?
- No. Detection and blocking are separate problems, and no registration system solves them. What Suede changes is what you hold when a clone surfaces: a timestamped, on-chain record of your reference recordings and published consent terms stating what use of your voice you have authorized. Platform takedowns and legal claims both start by asking for exactly that evidence. Registration means you have it dated before the incident instead of assembling it after.
- Is my voice protected by copyright?
- Your recordings are; your voice itself is not. Copyright covers the specific master recordings and compositions you create. The sound of your voice, as an attribute of your identity, falls under right of publicity, which in the United States is state law. Tennessee's ELVIS Act, passed in 2024, explicitly extended that protection to unauthorized AI simulation of a person's voice. Registering recordings on Suede does not create new legal rights; it creates the dated evidence and machine-readable terms that make asserting your existing rights practical.
- Can I license my voice to AI platforms on my own terms?
- Yes, and that is the other half of the point. Refusing all AI use is one valid position; licensing your voice model for uses you approve, at a price you set, is another. A registered voice asset carries license terms and a royalty route, so a platform or an agent that wants an authorized clone can read the terms, pay the contract, and the revenue splits to you and your collaborators automatically. Consent and compensation live in the same record.
Build your evidence first
Put your voice on the record.
Register your recordings on Suede to get dated, on-chain proof of ownership and consent terms that platforms, agents, and courts can check.
About the author
Jason Colapietro(Johnny Suede) is the founder and CEO of Suede Labs AI. He built the creator-ownership layer for the AI media era: proof of creation, programmable IP, on-chain royalty routing, and agent-accessible licensing. Patent pending USPTO 63/947,120.