v3.8.5

What's Changed

• fix: pin torchvision and torchcodec for torch 2.8 compatibility by @Barabazs in #1397

Full Changelog: v3.8.4...v3.8.5

v3.8.4

What's Changed

• feat: add progress_callback to transcribe, align, and diarize by @Barabazs in #1371
• fix: remove dead model_bytes read that leaked file handle by @Barabazs in #1381
• fix: restore word-level timestamps for unalignable characters by @Barabazs in #1386
• fix: require faster-whisper>=1.2.0 for use_auth_token support (#1385) by @Barabazs in #1388

Full Changelog: v3.8.2...v3.8.4

v3.7.9

Backport of word-level timestamp fixes from v3.8.4.

Bug Fixes

• Restore timestamps for unalignable characters (39aa9f5): Words containing digits, symbols, or foreign script (e.g. 4,9, £13.60) now get proper timestamps via a wildcard emission column. The previous patch (v3.X.Y) reverted PR #986 which removed wildcard support entirely. Fixes #1372.

Testing

• Add regression test for #1372 (da072d6)
• Add pytest dev dependency and CI test workflow (f9a3f8f)

Full Changelog: v3.7.8...v3.7.9

v3.6.2

Backport of word-level timestamp fixes from v3.8.4.

Bug Fixes

• Restore timestamps for unalignable characters (39aa9f5): Words containing digits, symbols, or foreign script (e.g. 4,9, £13.60) now get proper timestamps via a wildcard emission column. The previous patch (v3.X.Y) reverted PR #986 which removed wildcard support entirely. Fixes #1372.

Testing

• Add regression test for #1372 (da072d6)
• Add pytest dev dependency and CI test workflow (f9a3f8f)

Full Changelog: v3.6.1...v3.6.2

v3.5.2

Backport of word-level timestamp fixes from v3.8.4.

Bug Fixes

• Restore timestamps for unalignable characters (39aa9f5): Words containing digits, symbols, or foreign script (e.g. 4,9, £13.60) now get proper timestamps via a wildcard emission column. The previous patch (v3.X.Y) reverted PR #986 which removed wildcard support entirely. Fixes #1372.

Testing

• Add regression test for #1372 (da072d6)
• Add pytest dev dependency and CI test workflow (f9a3f8f)

Full Changelog: v3.5.1...v3.5.2

v3.4.5

Backport of word-level timestamp fixes from v3.8.4.

Bug Fixes

• Restore timestamps for unalignable characters (39aa9f5): Words containing digits, symbols, or foreign script (e.g. 4,9, £13.60) now get proper timestamps via a wildcard emission column. The previous patch (v3.X.Y) reverted PR #986 which removed wildcard support entirely. Fixes #1372.

Testing

• Add regression test for #1372 (da072d6)
• Add pytest dev dependency and CI test workflow (f9a3f8f)

Full Changelog: v3.4.4...v3.4.5

v3.3.6

Backport of word-level timestamp fixes from v3.8.4.

Bug Fixes

• Restore timestamps for unalignable characters (39aa9f5): Words containing digits, symbols, or foreign script (e.g. 4,9, £13.60) now get proper timestamps via a wildcard emission column. The previous patch reverted PR #986 which removed wildcard support entirely. Fixes #1372.

Testing

• Add regression test for #1372 (da072d6)
• Add pytest dev dependency and CI test workflow (f9a3f8f)

Full Changelog: v3.3.5...v3.3.6

v3.8.2

What's Changed

• feat: expose avg_logprob per segment from ctranslate2 beam search by @Barabazs in #1350
• fix: revert #986 wildcard alignment that broke word-level timestamps (#1220) by @Barabazs in #1367

Full Changelog: v3.8.1...v3.8.2

v3.7.8

Backport of word-level timestamp fixes from v3.8.2.

Bug Fixes

• Restore original CTC forced-alignment (f2609a6): PR #986 caused all words to anchor to the start of the segment window (silence) instead of actual speech. Reverts get_trellis/backtrack to the original PyTorch tutorial implementation. Fixes #1220.
• Fix blank_id hardcoded to 0 (636f298): Broke alignment for HuggingFace models where blank is [pad], not index 0.

Full Changelog: v3.7.7...v3.7.8

v3.6.1

Backport of word-level timestamp fixes from v3.8.2.

Bug Fixes

• Restore original CTC forced-alignment (f2609a6): PR #986 caused all words to anchor to the start of the segment window (silence) instead of actual speech. Reverts get_trellis/backtrack to the original PyTorch tutorial implementation. Fixes #1220.
• Fix blank_id hardcoded to 0 (636f298): Broke alignment for HuggingFace models where blank is [pad], not index 0.

Full Changelog: v3.6.0...v3.6.1