Audio

Speech-to-text (transcription) and text-to-speech, OpenAI-compatible.

Transcription (speech-to-text)

POST https://gateway.primex.work/v1/audio/transcriptions

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

Model	Notes	Billing
whisper	General-purpose STT	per audio minute ($0.017)
gpt-4o-transcribe	High-accuracy	per token
gpt-4o-transcribe-diarize	Adds speaker labels	per token
gpt-4o-mini-transcribe	Fast, economical	per token

Text-to-speech

POST https://gateway.primex.work/v1/audio/speech

curl https://gateway.primex.work/v1/audio/speech \
  -H "Authorization: Bearer prx-YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4o-mini-tts",
    "input": "Welcome to Primex.",
    "voice": "alloy"
  }' --output speech.mp3

Model	Billing
gpt-4o-mini-tts	per 1M characters

The audio route reads the requested model and routes accordingly, so the same endpoint serves every audio model in the catalogue.