Audio API

OpenAI-compatible audio endpoints for text-to-speech, speech-to-text transcription, and translation to English.

All endpoints share the base path /api/client/v1 and require a Bearer API token (cpeer_...).

Speech (Text-to-Speech)

Synthesizes spoken audio from input text. Returns raw audio bytes.

Endpoint

POST /api/client/v1/audio/speech

Request

Accepts a JSON body (Content-Type: application/json).

{
  "model": "tts-1",
  "input": "The quick brown fox jumped over the lazy dog.",
  "voice": "alloy",
  "response_format": "mp3",
  "speed": 1.0,
  "instructions": "Speak in a calm, measured tone."
}

Field	Type	Required	Description
model	string	Yes	TTS model key
input	string	Yes	Text to synthesize
voice	string	No	Voice name. If omitted, the provider falls back to its default voice
response_format	string	No	One of mp3, opus, aac, flac, wav, pcm. Invalid values are ignored and the provider default is used
speed	number	No	Playback speed multiplier
instructions	string	No	Free-form delivery/style instructions

Response

Raw audio bytes. The Content-Type header reflects the produced audio format, and the response includes:

• Content-Length — byte length of the audio
• X-Request-Id — request correlation ID

Example

curl -X POST https://gateway.example.com/api/client/v1/audio/speech \
  -H "Authorization: Bearer cpeer_your_token" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "tts-1",
    "input": "Hello world",
    "voice": "alloy",
    "response_format": "mp3"
  }' \
  --output speech.mp3

Transcriptions

Transcribes audio into text in the source language.

Endpoint

POST /api/client/v1/audio/transcriptions

Request

Accepts either multipart/form-data (file upload) or application/json (base64 audio).

Multipart form fields:

Field	Type	Required	Description
model	string	Yes	STT model key
file	file	Yes	Audio file to transcribe
language	string	No	Source language hint (e.g. en)
prompt	string	No	Optional text to guide the model
response_format	string	No	Transcript format (e.g. json, text, verbose_json)
temperature	number	No	Sampling temperature
timestamp_granularities[]	string (repeatable)	No	One or more of word, segment

JSON body uses the same fields, with the file supplied as a base64 audio object:

{
  "model": "whisper-1",
  "audio": {
    "data": "<base64-encoded audio>",
    "fileName": "speech.mp3",
    "contentType": "audio/mpeg"
  },
  "language": "en",
  "prompt": "",
  "response_format": "json",
  "temperature": 0,
  "timestamp_granularities": ["segment"]
}

Response

JSON containing the transcribed text and a request_id:

{
  "text": "The quick brown fox jumped over the lazy dog.",
  "request_id": "req_abc123"
}

The exact shape depends on response_format (e.g. verbose_json adds segment/word detail).

Example

curl -X POST https://gateway.example.com/api/client/v1/audio/transcriptions \
  -H "Authorization: Bearer cpeer_your_token" \
  -F model="whisper-1" \
  -F file="@speech.mp3" \
  -F response_format="json"

Translations

Transcribes audio and translates it into English. Same input handling as transcriptions, but language and timestamp_granularities[] are not used.

Endpoint

POST /api/client/v1/audio/translations

Request

Accepts either multipart/form-data (file upload) or application/json (base64 audio).

Multipart form fields:

Field	Type	Required	Description
model	string	Yes	STT model key
file	file	Yes	Audio file to translate
prompt	string	No	Optional text to guide the model
response_format	string	No	Transcript format (e.g. json, text, verbose_json)
temperature	number	No	Sampling temperature

The JSON body form mirrors transcriptions, supplying the audio as a base64 audio object.

Response

JSON containing the English text and a request_id:

{
  "text": "The quick brown fox jumped over the lazy dog.",
  "request_id": "req_abc123"
}

Example

curl -X POST https://gateway.example.com/api/client/v1/audio/translations \
  -H "Authorization: Bearer cpeer_your_token" \
  -F model="whisper-1" \
  -F file="@speech_fr.mp3" \
  -F response_format="json"

Provider Notes

Audio requests are routed to the underlying provider by the model key. For Azure-backed providers, audio endpoints are resolved by deployment rather than by model name, so a deployment must exist for the requested capability — for example, a cluster without a configured TTS deployment cannot serve /audio/speech. Ensure the target model maps to a deployment that supports the requested audio operation.

Errors

Status	Description
400	Missing/invalid required fields (model, input, file/audio.data) or unsupported Content-Type
401	Invalid API token
429	Rate limit, budget, or per-request quota exceeded
500	Inference error
503	Service is shutting down