//opengauntlet
// voices · what you're allowed to do

The licence traps, and the law that just landed

Open-weights speech models are unusually treacherous to license, four incompatible labelling regimes are now in force worldwide, and the market ships almost none of what they require. Informational only — this is not legal advice. Verify against primary sources and qualified counsel before relying on any of it.

The single most common mistake in open TTS

A speech project licenses its code and its model weights separately. MIT or Apache-2.0 code paired with non-commercial weights is extremely common, because the weights inherit restrictions from whatever the model was trained on.

The GitHub badge tells you about the code. Check the model card, and check it for the specific checkpoint you intend to ship — sibling models in one repository sometimes carry different licences. Three claims you will see everywhere and that are all wrong: Fish Audio S2 Pro is not MIT, IndexTTS 2 is not Apache-2.0, and F5-TTS's released checkpoints are not MIT.

Split licences, and other traps

  • Permissive code, restricted weights. The endemic pattern. Reading only the repository LICENSE file gives you the wrong answer for F5-TTS, OmniVoice, Echo-TTS, Supertonic and the Coqui/XTTS lineage.
  • Licences that cover the OUTPUT. A handful — the Coqui Public Model Licence most notably — restrict not just the model but the audio it generates. That is what makes XTTS-v2 unusable in a commercial product despite its popularity.
  • Open in name only. MegaTTS 3 is nominally Apache-2.0, but its publisher withholds the encoder weights, so you can only run inference from latents they provide. Voice cloning with your own audio is impossible.
  • Stacked licences. Some models inherit three at once — a project licence, a backbone-model licence, and a separate licence on the audio codec. All three have to clear.
  • Relicensing. Piper moved from MIT to GPL-3.0 and archived the old repository. Anyone who embedded the MIT-era build in a proprietary product cannot simply upgrade.
  • Transitive copyleft. eSpeak NG is GPL-3.0 and is the pronunciation front end embedded inside Piper, Kokoro, StyleTTS 2 and many others. Products shipping those models should check whether they are linking GPL code.

Four questions before you ship

  1. What licence is on the specific checkpoint file you will load — not the repository, the file?
  2. Does the licence say anything about the generated audio, as distinct from the model?
  3. What was it trained on, and does that corpus impose downstream terms?
  4. If it is a hosted API: who owns the output, is your audio used for training by default, how long is it retained, and what happens to a cloned voice if you stop paying?

Synthetic-audio labelling is now law in four places, and they disagree

European Union. AI Act Article 50 has applied since 2 August 2026. Providers of systems generating synthetic audio must mark outputs in a machine-readable format detectable as artificially generated; deployers publishing a deepfake must disclose it perceivably. A cloned voice that plausibly appears authentic counts, and the duty applies even without intent to deceive. Penalties reach €15 million or 3% of worldwide annual turnover. The regulation names no technical standard, so vendors are guessing between provenance metadata and watermarking.

China has the strictest and most technically prescriptive regime: a human-perceptible label and machine-readable metadata to a named national standard, in force since September 2025. India requires audio disclosures plus permanent provenance metadata, with takedown clocks measured in hours. South Korea requires visible or machine-readable labels for synthetic sounds. The UK, US federal law and Japan have no synthetic-audio labelling mandate at all.

The practical consequence: a single global product cannot satisfy all four in-force regimes with one implementation.

Almost nothing on the market is watermarked

Against those requirements, the market's actual behaviour is stark. Most major commercial vendors disclose no audio watermark. One leading vendor's own detection classifier explicitly does not reliably identify audio from its own flagship model, and analyses only the first minute of a clip.

The most reliably watermarked synthetic speech available right now comes from an MIT-licensed open model. Chatterbox embeds a watermark in every generation and ships the extraction code so anyone can verify it. For everything else, the practical retrofit is an open watermarking library with permissive weights, applied as a post-generation pipeline step.

Two caveats that apply to every scheme. All of them are proactive — they mark at generation time, so the absence of a watermark proves nothing. And published robustness figures are measured against benign edits like compression and re-encoding, not against a motivated adversary.

What it takes to clone a voice, by vendor

This varies enormously, and it is a procurement question rather than a feature. At the strict end, one hyperscaler requires a recorded verbal consent statement in a fixed script plus a biometric match between that statement and the training audio, and gates the whole capability behind account-team approval. In the middle, vendors require the speaker to read a consent statement with no verification that it is the same person. At the loose end sit vendors that accept a checkbox — and open-weights models, which have no gate whatsoever.

If deepfake liability matters to you, this dimension should outrank price. And note that self-hosting exempts you from nothing: if you place a self-hosted system on the EU market, you are the provider under Article 50.

Does your vendor train on your audio?

Several major TTS vendors train on customer content by default, with an opt-out buried in account settings that is not retroactive. At least two take a perpetual, irrevocable, sublicensable licence over your input with no opt-out identified at all — difficult to reconcile with data-subject rights under GDPR.

Others do not train by default and say so plainly; one exposes a per-request parameter to disable it, which is the most auditable control found anywhere in the segment. Retention and training policy could not be verified from primary sources for most vendors. Speech inputs routinely contain personal data — get these terms contractually rather than assuming.

US voice rights, briefly

Copyright does not protect a voice. The leading 2025 case held that copyright covers the fixed sound recording, not the abstract qualities of a voice nor a new recording that imitates one. Trademark generally fails too. What survives is state right-of-publicity law, state consumer-protection law, and breach of contract — and the question of whether training on scraped recordings infringes was left open rather than decided.

The fastest-moving US exposure is not publicity but biometric privacy. Illinois's statute enumerates voiceprints and carries a private right of action; a wave of class actions filed in May 2026 named nine major technology companies over voiceprints allegedly extracted from public recordings to train commercial voice models. The unresolved doctrinal question — whether a generative voice model is a "voiceprint" — will determine exposure for the entire industry.

Cloning the voices of the dead

There is no binding law anywhere governing memorial voices or digital-afterlife services. The only articulated guidance is academic: consent from the person before death, consent from heirs and from the mourners who will interact, adult users only, easy opt-out, and defined retirement protocols for a dignified shutdown.

The nearest binding constraints are indirect — post-mortem publicity rights, and estate consent under performers'-union agreements. In most of Europe, data-protection law does not apply to the deceased at all. The result is an inversion worth stating plainly: the law protects the commercial value of a dead celebrity's voice far better than it protects an ordinary family from an unwanted synthetic parent.

A checklist

  1. If you touch the EU, implement machine-readable marking on output. An MIT-licensed watermarking library with open weights is the free path.
  2. Give whoever deploys your system a perceivable disclosure mechanism — that duty is theirs, but they need a hook from you.
  3. If you clone voices, copy the strictest vendor pattern: fixed-script recorded consent plus a biometric match against the training audio, retained as evidence.
  4. Never rely on "the user ticked a box" as consent evidence in a biometric-privacy jurisdiction.
  5. Document your training-data provenance now — that is exactly what the 2026 litigation wave targets.
  6. Check whether your vendor trains on your audio by default, and turn it off in writing.
  7. For deceased voices, obtain estate consent and treat the academic guidance as the floor.
  8. Do not assume any vendor watermarks unless its documentation says so. Most do not.