TRANSCRIPTION_AUDIO

Audio Transcription

• Supported providers:
  • OpenAI (OPENAI_API_KEY + OPENAI_API_TRANSCRIBE_MODEL)
  • Groq (OpenAI-compatible) — recommended as a cost-effective alternative

Groq Configuration

• Environment variables:
  • GROQ_API_KEY: your Groq API key
  • GROQ_API_TRANSCRIBE_MODEL (optional): defaults to whisper-large-v3
  • GROQ_API_BASE_URL (optional): defaults to https://api.groq.com/openai/v1

OpenAI Configuration

• Environment variables:
  • OPENAI_API_KEY: your OpenAI API key
  • OPENAI_API_TRANSCRIBE_MODEL: e.g. gpt-4o-mini-transcribe or whisper-1

Priority

• If GROQ_API_KEY is set for the session, transcription uses Groq.
• Else, if OPENAI_API_KEY is set, it uses OpenAI.
• If neither is set, it falls back to local Whisper via audio2textjs.

Per-session overrides

• When using Redis-backed configs, you can define per-session values which override environment defaults. These values can be set through the web UI modal or via the unoapi-config template.
• Per-session fields:
  • groqApiKey, groqApiTranscribeModel, groqApiBaseUrl
  • openaiApiKey, openaiApiTranscribeModel

Endpoint Note

• The Groq endpoint follows OpenAI compatibility: POST /audio/transcriptions with multipart form (file, model).