All pull requests

Pull requests

• Feature/windows portable release

  #1494·aniki0220 opened Oct 20, 2026·

  ·

• feat: optional input normalization for Hugging Face align models

  #1493·Damikk19 opened Oct 8, 2026·

  ·

• fix: ValueError when transcribe() language differs from load_model()

  #1492·Damikk19 opened Oct 8, 2026·

  ·
  1

• fix(subs): keep a subtitle line that fits in the character limit

  #1488·SashaMIT opened Sep 29, 2026·

  ·

• fix: start a new subtitle cue when the speaker changes

  #1487·MohammadHijjawi97 opened Sep 28, 2026·

  ·

• feat: add --num_speakers CLI flag for a known speaker count

  #1486·sighonara opened Sep 28, 2026·

  ·
  1

• fix(vad): restore torch thread count after loading Silero VAD

  #1484·jardelapp opened Sep 22, 2026·

  ·

• feat: auto-detect Ascend NPU and other non-CUDA accelerators

  #1483·li-lizhe opened Sep 21, 2026·

  ·
  2

• Fix/empty segments handling

  #1480·hsynylmz-umhg opened Sep 14, 2026·

  ·

• fix(audio,alignment): add os.PathLike support for audio file inputs

  #1479·hsynylmz-umhg opened Sep 14, 2026·

  ·

• Add --diarize_device so diarization can run on a different device than ASR

  #1476·ces3001 opened Sep 7, 2026·

  ·

• Sync optional torchaudio change with uv lockfile and alignment fallback

  #1473·m-bain with CopilotAI opened Aug 30, 2026·

  ·
  

• fix(cli): write the transcribed language instead of the --language default

  #1472·Kayvan-Zahiri opened Aug 28, 2026·

  ·

• Make torchaudio optional with Hugging Face fallback for align models

  #1469·li-lizhe opened Aug 27, 2026·

  ·
  11

• Split Chinese and Japanese into sentences during alignment

  #1466·Kayvan-Zahiri opened Aug 13, 2026·

  ·
  1

  2

• Specify master branch for loading Silero via torch hub

  #1465·kingjan1999 opened Aug 12, 2026·

  ·

• Format transcription timestamp as HH:MM:SS.sss

  #1464·JonstonChan opened Aug 5, 2026·

  ·

• fix(cli): read the Hugging Face token from HF_TOKEN (#1457)

  #1460·Anai-Guo opened Jul 30, 2026·

  ·
  1

• feat: add use_tmp_file option to load_audio for low-memory decoding

  #1455·katherineahn opened Jul 26, 2026·

  ·
  1

• Batch forced-alignment inference

  #1454·tlebryk opened Jul 22, 2026·

  ·

• feat: Optimize alignment post-processing with NumPy

  #1453·tlebryk opened Jul 17, 2026·

  ·

• feat: expose optional diarization overlap and exclusive outputs

  #1447·numsu opened Jul 1, 2026·

  ·

• feat: add time_weighted interpolation and --timestamp_sanitize flag

  #1444·hiyochi opened Jun 30, 2026·

  ·

• feat: add --highlight_color and --highlight_score_colors options for word-level highlight (#415)

  #1443·hiyochi opened Jun 30, 2026·

  ·
  1

• Gracefully skip alignment for languages with no align model (#298)

  #1442·MushiSenpai opened Jun 27, 2026·

  ·
  1