Back to Industries
Music & Voice

A 30-second sample clones a voice. The industry has no guardrails for it.

Labels and studios you work with aren’t the problem — those relationships have contracts. The threat is outside: AI companies generating songs in your style from public recordings, overseas platforms dubbing your voice into languages you never authorized, synthetic replacements trained by entities that never signed a deal. Seedance-style tools can generate convincing audio from a 30-second sample. The industry has no shared infrastructure for voice-level consent — no guardrails for artists to control authorization, and no way for labels or platforms to prove they had it. DIAP gives everyone the tools: Voice Module consent, Sound IDs, ultrasonic watermarks, and a permanent provenance trail.

30s

Sample needed to clone a voice

40+

Languages AI can dub into

0

Artist consent mechanisms

What DIAP Does For Music & Voice

Voice Module Consent

A voice is registered as a separate identity module with independent authorization controls. Authorizing face rendering does not authorize voice cloning. Every module requires explicit, separate consent.

TRAINING_USE Rights Flag

A dedicated rights flag that must be explicitly granted before a voice can be used to train any AI model. Without it, voice data cannot legally enter any training pipeline under DIAP authorization.

Sound IDs & Ultrasonic Watermarks

Every authorized audio output gets a unique Sound ID and an ultrasonic watermark embedded in an inaudible frequency range — detectable by DIAP’s verification system. Audio verification works without the original file.

ScriptModule — For Songwriters & Lyricists

Lyrics, compositions, and written works get the same protection as voice. ScriptModule registers written works with a cryptographic fingerprint. SCRIPT_TRAINING_USE is a separate right — a label or platform requesting lyrics for analysis does not get the right to train a model on them. AuthorshipReceipts log every access permanently.

Emergency Revocation — Shared Safety Mechanism

Emergency revocation blocks unauthorized voice cloning and AI audio generation — it never affects valid licenses. If a label has a two-year dubbing authorization, that deal is honored. It targets the threats outside valid agreements: the AI company cloning a voice without a contract, the foreign platform generating songs without authorization. Anyone in the authorized chain — artists, agents, labels, guild reps — can trigger it when they detect unauthorized use.

Build On Our Infrastructure

Labels, distributors, and music platforms can integrate DIAP directly into their own ecosystems using the open API.

How It Works For Music & Voice

1

Register a voice module

Self-service via phone mic or studio-quality recording. The voice is converted to a one-way audio embedding. The raw recording is discarded after extraction.

2

Set per-project authorization

Each dubbing project, AI song generation, or voice synthesis request requires a separate token. You specify which rights are granted: VOICE_RENDER, VOICE_DUBBING, VOICE_CLONE — and explicitly whether TRAINING_USE is permitted.

3

Monitor and verify

Every authorized audio output carries a Sound ID. DIAP’s blind audio verification (blind audio frequency analysis) can detect your watermark in any audio file without the original. Unauthorized clones are detectable.

Real Scenario

Artist authorizes English dubbing only. AI attempts Spanish clone. Denied.

1

Voice Module registered, VOICE_DUBBING authorized for English only

2

English dubbing token issued, 3 render receipts submitted

3

Spanish dubbing request submitted — territory not in token scope

4

TRAINING_USE flag not set — AI training request denied automatically

5

Artist sees both denied requests in their audit trail

What Music & Voice Artists Get

  • Voice Module with independent authorization controls
  • TRAINING_USE rights flag — explicit opt-in required
  • Sound IDs on every authorized audio output
  • Ultrasonic watermarks — inaudible, detectable
  • Blind audio verification via blind audio frequency analysis
  • ScriptModule for songwriters — lyrics and compositions protected
  • Emergency revocation — blocks unauthorized use without affecting valid licenses
  • Open API for labels and distributors to integrate directly
  • Free for artists — always