Skip to content

audio-metadata.md

Latest commit

 

History

History
39 lines (25 loc) · 1.7 KB

audio-metadata.md

File metadata and controls

39 lines (25 loc) · 1.7 KB
Raw

Feature: Audio Metadata Extraction

Purpose

Read duration, sample rate, channels, bit depth, and codec from the source audio archived at the end of each live event (events/<id>/source.<wav|ogg|opus>). Pure-Python: stdlib wave for uncompressed WAV, mutagen for everything else. No ffmpeg dependency.

When it runs

Post-event, not during the live stream. When the speaker disconnects and the source recording is finalized, the service layer calls service.audio_metadata.extract_metadata() and stamps the result onto the event manifest (duration_ms field) so the dashboard's "Interpretation Minutes" tile and the per-event detail view both have accurate numbers without re-reading the audio bytes on every request.

Inputs

  • Raw bytes of the source recording (WAV 16-bit / 24 kHz mono written by service.audio_archive from the speaker's PCM stream).
  • Original filename / suggested extension.
  • Declared content type.

Outputs

FileMetadataDetail (defined in services/api/app/types/files.py):

  • duration_ms: int | None
  • sample_rate: int | None
  • channels: int | None
  • bit_depth: int | None
  • codec: str | None

Any value is allowed to be null — partial metadata is better than failing the event finalization because of an unusual encoder.

Edge cases

  • Mid-event abort with no recording yet — finalization skips the extraction step entirely; the manifest carries duration_ms = null.
  • Unsupported container — mutagen raises; the extractor swallows the exception, logs a warning, and returns nulls.
  • Corrupt header — same handling.

Related