v0.27.3

[eginhard]

What's Changed

Features

• Return sentence level timestamps for multi-sentence input by @santosh-r24 in Feature: Sentence level timestamps #520 (documentation)

Fixes

• Support Pytorch 2.9 by @eginhard in fix: support pytorch 2.9 #525
• fix(xtts): update streaming code to transformers>=4.57 by @eginhard in fix(xtts): update streaming code to transformers>=4.57 #505
• fix(xtts): add over limit text display by @zkhan122 in Add over limit text display #495
• Handle romanization for Fairseq models by @eginhard in Handle romanization for Fairseq models #506
• docs(formatters): show expected metadata format, improve error handling by @eginhard in docs(formatters): show expected metadata format, improve error handling #511 (documentation)
• fix(bark): ensure tensors are on same device when not using reference audio by @eginhard in fix(bark): ensure tensors are on same device when not using reference audio #512
• fix(audio): create parent folders, improve error message by @eginhard in fix(audio): create parent folders, improve error message #513
• fix(docker): correct torch backend selection by @eginhard in fix(docker): correct torch backend selection #524
• Replace jieba with spacy_pkuseg for Chinese word segmentation by @langfod in Replace jieba with spacy_pkuseg #504

New Contributors

• @langfod made their first contribution in Replace jieba with spacy_pkuseg #504
• @zkhan122 made their first contribution in Add over limit text display #495
• @santosh-r24 made their first contribution in Feature: Sentence level timestamps #520

Full Changelog: v0.27.2...v0.27.3

---

This discussion was created from the release v0.27.3.