Audio

Turn text into speech using a cloned or preset voice, then poll for the finished track.

Generate speech (text-to-speech) from a voice you own or a system preset.

Create audio

POST /v1/audio

Asynchronous — returns 201 with status: "processing". Requires an Idempotency-Key.

Field	Type	Required	Default	Description
`voice_id`	string (UUID)	yes	—	A ready voice you own, or a preset
`text`	string	yes	—	1–5000 characters
`name`	string	no	`Audio YYYY-MM-DD`	Track name (max 200)
`model`	enum	no	voice’s model, else `minimax`	`minimax` or `qwen3`
`emotion`	enum	no	`auto`	`auto`, `neutral`, `happy`, `sad`, `angry`, `fearful`, `disgusted`, `surprised`
`webhook_url`	string	no	—	Held — returns `503 webhook_delivery_unavailable`

Qwen3-specific (optional): speaker, voice_description, style_instruction, reference_audio (URL, SSRF-validated), reference_text, language.

{
  "data": {
    "id": "track_uuid",
    "name": "Audio 2026-03-29",
    "status": "processing",
    "credits_used": 3,
    "credits_remaining": 247,
    "created_at": "..."
  }
}

List audio tracks

GET /v1/audio

Paginated. Query params: voice_id, cursor, limit.

Get an audio track

GET /v1/audio/:id

Poll until status is ready or failed. When ready, data.audio_url holds the CDN URL.

Delete an audio track

DELETE /v1/audio/:id

Deletes the track.