0.1.4

elyxlz/voxtral0.1.4Apr 13, 2026by zhu-han

AI Summary

Fixes inference batch processing issues and switches audio libraries for better device compatibility.

Key Highlights

Fix infer_batch instruction
Batch inference without ref_text
Switch to soundfile+librosa for device compatibility
Google Colab examples added

New Features

Robust batch inference
Cross-platform audio handling

Full Release Notes

## What's Changed
* fix: instruct in infer_batch by @Pastells in https://github.com/k2-fsa/OmniVoice/pull/72
* docs: add omnivoice-server to community projects by @maemreyo in https://github.com/k2-fsa/OmniVoice/pull/42
* fix: batch_inference without ref_text or ref_audio_path by @Pastells in https://github.com/k2-fsa/OmniVoice/pull/70
* docs: restore omnivoice-server to community projects (after force-push) by @maemreyo in https://github.com/k2-fsa/OmniVoice/pull/80
* fix infer_batch.py for mixed modes https://github.com/k2-fsa/OmniVoice/commit/58cf379ad4097eaa03e5d6e9d361e08bbd4074bb
* use soundfile+librosa instead of torchaudio to avoid issues on some devices https://github.com/k2-fsa/OmniVoice/commit/4f4b0ccb6e103daf810608d0ecb0e18a0d8b2cde
* replace Chinese parentheses with English ones in text https://github.com/k2-fsa/OmniVoice/commit/31b8a4d8b7c5e804644a774f12f665234663b6fa
* add more tips in README https://github.com/k2-fsa/OmniVoice/commit/8ede7acd2af94b810320091aad4e307d259e9690
* add google colab examples https://github.com/k2-fsa/OmniVoice/commit/936146629b1fa91d2391db8aa56d49b3e4510070

## New Contributors
* @Pastells made their first contribution in https://github.com/k2-fsa/OmniVoice/pull/72
* @maemreyo made their first contribution in https://github.com/k2-fsa/OmniVoice/pull/42

**Full Changelog**: https://github.com/k2-fsa/OmniVoice/compare/0.1.3...0.1.4