CN8 API Documentation

Two cloning workflows — persistent (async, reusable voice_id) and instant (sync, returns binary audio).

Persistent clone (core-voice-clone) is a delayed service: 202 + jobid; poll /v2/jobs/{jobid} for the voice_id. Instant clone (core-instant-clone-tts) is sync but returns BINARY audio/wav with metadata in response headers. Both accept audioUrl (camelCase) for the source sample (mp3/wav/ogg or a video — audio is extracted via ffmpeg).

• Use upload-media to upload local audio/video and get a URL.
• Persistent: cheapest per call when reusing the voice (one-time $5).
• Instant: cheapest for one-off clips ($2/request, raw audio response).
• Cloning is provider-scoped — voice_id from elevenlabs cannot be used with minimax/openai.