openai-whisper

This project publishes no release notes. Between v20240930 and v20250625 there were 20 commits, 18 of them substantive:

• more pytorch versions in tests (#2408)
• Add option to carry initial_prompt with the sliding window (#2343)
• Update README.md (#2379)
• fix typo data/README.md (#2433)
• Updating README and doc strings to reflect that n_mels can now be 128 (#2049)
• Bugfix: Illogical "Avoid computing higher temperatures on no_speech" (#1903)
• Upgrade GitHub Actions (#2430)
• pre-commit autoupdate && pre-commit run --all-files (#2484)
• PEP 621: Migrate from setup.py to pyproject.toml (#2435)
• Update python-publish.yml
• GitHub Actions: Add Python 3.13 to the testing (#2487)
• pre-commit: Upgrade black v25.1.0 and isort v6.0.0 (#2514)
• Keep GitHub Actions up to date with GitHub's Dependabot (#2486)
• Bump the github-actions group with 3 updates (#2592)
• Fix: GitHub display errors for Jupyter notebooks (#2589)
• Fixed triton kernel update to support latest triton versions (#2588)
• Fix: Ensure DTW cost tensor is on the same device as input tensor (#2561)
• Fix: Update torch.load to use weights_only=True to prevent security w… (#2451)

This project publishes no release notes. Between v20240927 and v20240930 there were 5 commits, 4 of them substantive:

• using sdpa if available (#2359)
• test on python/pytorch versions up to 3.12 and 2.4.1 (#2360)
• large-v3-turbo model (#2361)
• allowing numpy 2 in tests (#2362)

This project publishes no release notes. Between v20231117 and v20240927 there were 5 commits, 4 of them substantive:

• Fix triton env marker (#1887)
• Skip silence around hallucinations (#1838)
• Relax triton requirements for compatibility with pytorch 2.4 and newer (#2307)
• pinning numpy<2 in tests (#2332)

This project publishes no release notes. Between v20231106 and v20231117 there were 2 commits, 1 of them substantive:

• Relax triton requirements for compatibility with pytorch 2.1 and newer (#1802)

This project publishes no release notes. Between v20231105 and v20231106 there were 2 commits, 1 of them substantive:

• large-v3 (#1761)

This project publishes no release notes. Between v20230918 and v20231105 there were 7 commits, 6 of them substantive:

• Fix exception when an audio file with no speech is provided (#1396)
• Add new option to generate subtitles by a specific number of words (#1729)
• handling transcribe exceptions. (#1682)
• allow_pickle=False while loading of mel matrix IN audio.py (#1511)
• docs: Disambiguation of the term "relative speed" in the README (#1751)
• remove tiktoken pin (#1759)

This project publishes no release notes. Between v20230314 and v20230918 there were 22 commits, 21 of them substantive:

• Update tokenizer.py (#1163)
• python-publish.yml: bump actions version to fix node warning (#1211)
• Squash long words at window and sentence boundaries. (#1114)
• Implement max line width and max line count, and make word highlighting optional (#1184)
• Update README.md to reference tiktoken (#1105)
• Update decoding.py (#1155)
• Update decoding.py (#1219)
• Python 3.11 (#1171)
• Drop ffmpeg-python dependency and call ffmpeg directly. (#1242)
• Dropped unused execute bit from mel_filters.npz. (#1254)
• Avoid computing higher temperatures on no_speech segments (#1279)
• Updated README.md to provide more insight on BLEU and specific appendices (#1236)
• Fix numba depreceation notice (#1233)
• fix condition_on_previous_text (#1224)
• Improve timestamp heuristics. (#1461)
• Avoid rearranging all caches (#1483)
• word timing tweaks (#1559)
• Update model-card.md (#1643)
• fix doc of TextDecoder (#1526)
• Add .pre-commit-config.yaml (#1528)

…and 1 more.

This project publishes no release notes. Between v20230308 and v20230314 there were 6 commits, 5 of them substantive:

• Use tiktoken (#1044)
• Fix alignment between the segments and the list of words (#1087)
• fix github language stats getting dominated by jupyter notebook (#1076)
• Fix truncated words list when the replacement character is decoded (#1089)
• abort find_alignment on empty input (#1090)

This project publishes no release notes. Between v20230307 and v20230308 there were 4 commits, 3 of them substantive:

• fix typo
• fix all_tokens handling that caused more repetitions and discrepancy in JSON (#1060)
• kwargs in decode() for convenience (#1061)