Multiple small improvements to transcription, with miscellaneous other fixes and usability improvements.
See the v1.7.0 release notes for full details on the transcription system.
v1.7.3 is also the first release that ships the PySubtrans library as a pip package, with support for integrating transcription in your own software.
What's Changed
- Recognise voice tags that wrap multi-line VTT cues by @machinewrapped in #492
- Recognise common titles when finding sentence ends by @machinewrapped in #491
- Keep squeezed words for speaker and sentence structure under partial coverage by @machinewrapped in #503
- Don't extend a word to opening punctuation after it by @machinewrapped in #506
- Time parts by zero-duration words by @machinewrapped in #507
- Added an option to place the translation above the original text by @machinewrapped in #508
- Fit subtitle row heights to the view width by @machinewrapped in #487
- Post-process selections that include untranslated lines by @machinewrapped in #488
- Offer OpenRouter diarization only for models that support it by @machinewrapped in #498
- Stop Gemini batches that stream for too long by @machinewrapped in #500
- Add Requesty to the readme translation providers by @Thibaultjaigu in #494
Full Changelog: v1.7.2...v1.7.3