Skip to contentATP Token
POST/omni/media/v1/audio/generations

Speech generation (TTS)

Text-to-speech is synchronous — text in, one audio file out as a signed URL. Same base URL and atp- key. The gateway routes your unified model to an audio provider (e.g. fish-s2-pro) and speaks the right dialect. There is no async task; GET /omni/media/v1/audio/tasks/{id} always returns 404.

  • Output is written to object storage and returned as a signed edge URL (https://media-<env>.atptoken.ai/v/...) with a 30-minute TTL — never inline base64. Fetch it promptly.
  • Requests are refused with 402 insufficient_quota when the project balance is ≤ 0.

Generate speech#

curl https://api.atptoken.ai/omni/media/v1/audio/generations \
  -H "Authorization: Bearer atp-..." -H "Content-Type: application/json" \
  -d '{ "model": "fish-s2-pro", "text": "Hello from ATP Token.", "format": "mp3" }'
FieldTypeDescription
modelstring · Requnified audio model (audio pool)
textstring · Reqtext to speak
reference_idstringvoice id (string, or array for multi-speaker)
formatstringmp3 (default) / wav / pcm / opus
sample_rateintegerHz (format default)
prosodyobject{ "speed": 1, "volume": 0, "normalize_loudness": true }
temperaturenumberexpressiveness 0–1 (default 0.7)

OpenAI-style clients: OpenAI TTS models also accept the OpenAI shape — input, voice, response_format — which the gateway maps.

Response#

200

{
  "created": 1781776187,
  "data": [
    { "url": "https://media-prod.atptoken.ai/v/audio/aud_....mp3?exp=...&sig=..." }
  ]
}

Errors#

400
missing `model` / `text`, or an invalid `format`.
402
`insufficient_quota`: project balance ≤ 0; top up and retry (don't hammer).
422
the model has no audio provider.

Next steps

PreviousVideo generationNextConsole API overview
Was this page helpful?