Skip to content

v0.8.1 — robust overlap finalization

Choose a tag to compare

View all tags
@apiad apiad released this 29 Jun 15:43
v0.8.1
d8be002

Fixes & improvements

Overlap finalization + segment-timestamp dedup. Fixed-window chunk cuts land mid-word, which garbled weak models and could duplicate a phrase at the commit seam. The engine now commits finalized chunks by absolute segment timestamp and holds back the trailing stream_overlap seconds (cut mid-word) to be re-decoded with full context in the next chunk.

Changed

  • base streaming is now robust and complete — recovers quiet tails a single batch decode dropped, and no seam duplication.
  • Small models (tiny) go from garbled/hallucinated output to mostly coherent on fluent/pause-less audio. (Weak models can still hallucinate sporadically — base+ recommended for production.)
  • Repetition hallucinations are dropped (segment compression ratio > 2.4, Whisper's own failed-decode signal).
  • New LocalWhisperEngine.transcribe_segments; re-added stream_overlap (default 3.0s).

See CHANGELOG.md.

Assets 2
Loading