deepgram-transcribe

Transcribe audio via Deepgram Nova-2.

API Key

Set env var: DEEPGRAM_API_KEY

Usage

Transcribe a local file

FILE=/path/to/audio.ogg

# content-type guess
EXT="${FILE##*.}"
case "$EXT" in
  ogg) CT="audio/ogg" ;;
  mp3) CT="audio/mpeg" ;;
  wav) CT="audio/wav" ;;
  m4a) CT="audio/mp4" ;;
  webm) CT="audio/webm" ;;
  *) CT="audio/ogg" ;;
esac

curl -s -X POST \
  "https://api.deepgram.com/v1/listen?model=nova-2&punctuate=true&smart_format=true" \
  -H "Authorization: Token $DEEPGRAM_API_KEY" \
  -H "Content-Type: $CT" \
  --data-binary @"$FILE" \
  | python3 -c "import sys,json; d=json.load(sys.stdin); print(d['results']['channels'][0]['alternatives'][0]['transcript'])"

Notes

• For non-English audio, add &language=xx (e.g. language=ru).
• If transcript is empty, try removing language to auto-detect.