Audio

Speech synthesis, transcription, and translation with the same auth and billing gates.

Speech

POST/v1/audio/speechAPI key

curl -X POST https://api.relayrouter.org/v1/audio/speech \
  -H "Authorization: Bearer $RELAY_API_KEY" \
  -H 'content-type: application/json' \
  -d '{
    "model": "tts-model",
    "input": "Welcome to RelayRouter.",
    "voice": "alloy"
  }' \
  --output speech.mp3

Returns audio bytes, not JSON. Billing uses input text tokens plus the network fee.

Transcription and translation

Endpoint	Body	Response
POST /v1/audio/transcriptions	multipart	`file`, `model`, optional `language`.
POST /v1/audio/translations	multipart	Same shape; output is translated to English.
POST /v1/transcripts	multipart	Alias of transcriptions.

curl -X POST https://api.relayrouter.org/v1/audio/transcriptions \
  -H "Authorization: Bearer $RELAY_API_KEY" \
  -F [email protected] \
  -F model=whisper-model \
  -F language=en

Returns { "text": "..." } for transcription endpoints.

Audio transcription pricing is estimated from upload bytes at roughly 2 MB per minute, with a minimum of one minute. If a model has no audio pricing configured, the call falls back to token billing on estimated audio tokens. The exact amount is recorded in usage_logs.cost_micro.

10 MB limit

Audio uploads are bounded by the global request body limit. For longer recordings, split the file into chunks before uploading.

PreviousImages NextVideo