Recipe: Transcribe audio, then translate

Chain two services: transcribe an audio file with speech-to-text, then translate the text into one or more languages. Both bill through your prx- key and charge only on success.

Prerequisites

• A prx- key (Getting Started).
• A speech-to-text model ID and a translation model ID from the live catalogue.
• An audio file (e.g. meeting.mp3).

1. Transcribe (speech-to-text)

curl https://gateway.primexos.com/v1/audio/transcriptions \
  -H "Authorization: Bearer prx-YOUR_KEY" \
  -F model="YOUR_STT_MODEL" \
  -F file="@meeting.mp3"

The response contains the transcript text. STT is billed per audio minute. See Audio.

2. Translate the transcript

curl https://gateway.primexos.com/v1/translate \
  -H "Authorization: Bearer prx-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "YOUR_TRANSLATION_MODEL",
    "text": "PASTE_TRANSCRIPT_HERE",
    "to": ["es", "fr"]
  }'

to accepts one language or several at once; omit from to auto-detect the source. Translation is billed per 1M characters. See Translation and Language and Safety.

Notes

• To avoid publishing the audio, upload it once through the Files API and pass the file_id instead of the raw file.
• Each stage is a separate, independently metered request; a failure in either costs $0. See Billing & Credits.