Skip to content

Muse Voice Transcribe

Transcribes voice notes from the web app and from the phone, kept on the item as the owner's words.

Built so far

  • Request. POST /v1/asr/transcribe, multipart: a request part (JSON with model muse-voice-transcribe-1.0, audioEncoding WAV, mode PUSH_TO_TALK) and the audio part. WAV only, ten minutes at most, 32 MB at most; the service reads the WAV header first and refuses anything else before a call is made (VOICE_MAX_SECONDS).
  • Response. transcript for the whole clip and audioDurationMs; priced by the seconds heard, recorded in the ledger under transcribe. Fixture mode answers from tests/fixtures/api/asr-transcribe/transcribe.json.
  • Where it lands. The item's condition_notes field with the source voice_note, accepted as the owner's words; see voice notes.