Voice Cloning API — CosyVoice3 & Qwen3-TTS on KeyoAPI

Need a voice cloning API without standing up a research stack? KeyoAPI exposes clone-friendly async TTS (CosyVoice3, Qwen3-TTS) behind one prepaid key — draft scripts with free chat twins, synthesize, then optionally animate with Duix or InfiniteTalk.

Get API key Open Model Square Full pricing list Free models

Clone-friendly speech models

Indicative figures for planning. Confirm live Keyo sell rates on interactive Model Square or the static pricing list.

Model ID	Fit	KeyoAPI	Notes
`Qwen3-TTS`	Async TTS with rapid voice cloning for brand / locale voices	~$0.548 / 10K characters	/pricing/Qwen3-TTS · guide /model/Qwen3-TTS
`CosyVoice3`	Async multilingual zero-shot speech; strong for longer narration	~$1.37 / 10K characters	/pricing/CosyVoice3 · guide /model/CosyVoice3
`IndexTTS-2`	Sync TTS when you need audio in one round trip (clone where schema allows)	~$1.37 / 10K characters	/pricing/IndexTTS-2 · guide /model/IndexTTS-2

Consent and abuse first

Clone samples only with rights clearance. Rate-limit tenant clones, reject celebrity likeness without licenses, and keep a default voice fallback. Keyo provides the inference API — your product owns consent UX and policy.

Rates and endpoints

Use it for brand-consistent voices across IVR, support bots, and avatar video. Keyo collocates clone-friendly TTS with Whisper STT and talking-avatar models so the full loop shares one balance.

Indicative rates above are catalog snapshots — confirm /pricing/Qwen3-TTS and /pricing/CosyVoice3. There is no separate upload playground: after /sign-up, try from the interactive pricing page for each model id.

How to call: async create via POST /v1/async/audio/speech with model=Qwen3-TTS or CosyVoice3, then poll /v1/task/{id}. Broader TTS map: /tts-api. Avatar handoff: /ai-avatar-video-generator.

For sync-only demos without a clone workflow, IndexTTS-2 on /v1/audio/speech is often enough.

curl https://www.keyoapi.xyz/v1/async/audio/speech \
  -H "Authorization: Bearer $KEYO_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model":"Qwen3-TTS","input":"Hello from a cloned brand voice."}'
# then: GET /v1/task/{id}

FAQ

Do you store my clone samples?

Follow Keyo docs and your own retention policy. Treat clone audio as sensitive PII; minimize retention in your app layer.

Voice cloning vs plain TTS?

Plain TTS uses stock voices; cloning targets a specific speaker sample. Prefer Qwen3-TTS or CosyVoice3 for clone workflows; see each model guide for schema details.

How do I try it?

Create a key at /sign-up, then open /pricing/Qwen3-TTS or /pricing/CosyVoice3. No second playground.

Also see /compare, /free-models, and model guides under /model/.