Starstruk Logo MarkStarstruk

2025

AudarAI/Audar-ASR-V12025Feb 6, 2026by andimarafioti

AI Summary

This release introduces support for multiple new speech models including ChatTTS, Moonshine, and Paraformer, along with macOS MPS acceleration and Docker support. It also adds REST API compatibility and refactors the codebase for better structure and stability.

Key Highlights

  • Added support for ChatTTS, Moonshine, and Paraformer (Chinese) models.
  • Introduced macOS MPS acceleration and Dockerization.
  • Added REST call support similar to OpenAPI style.
  • Implemented DeepFilterNet for speech enhancement.
  • Refactored handlers and arguments folder for improved code quality.

Breaking Changes

  • Changed keras backend to torch.
  • Changed default value of audio_enhancement from [True] to [False].

New Features

  • ChatTTS support with Chinese capabilities
  • Moonshine model integration
  • Paraformer ASR for Chinese speech recognition
  • DeepFilterNet for speech enhancement
  • macOS MPS acceleration support
  • Docker containerization
  • REST API endpoint support
  • Multi-language support

Full Release Notes

## What's Changed
* Minor doc fix. by @Vaibhavs10 in https://github.com/huggingface/speech-to-speech/pull/2
* Fix missing sounddevice module by @AlexHayton in https://github.com/huggingface/speech-to-speech/pull/7
* Update README.md by @RodriMora in https://github.com/huggingface/speech-to-speech/pull/23
* fix issue with ntlk by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/29
* Dockerize by @codearranger in https://github.com/huggingface/speech-to-speech/pull/22
* Add support to MPS by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/20
* adding apache license by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/31
* refactor arguments folder + run ruff by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/32
* Allow LM selection and MLX Gemma by @RonanKMcGovern in https://github.com/huggingface/speech-to-speech/pull/40
* Improvements mlx pipeline by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/41
* refactor all the handlers - folder structure by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/43
* add min new tokens by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/49
* add warning to install flash attn by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/51
* improve logging by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/52
* feat:add paraformer_zh asr by @wuhongsheng in https://github.com/huggingface/speech-to-speech/pull/48
* Assigning min new tokens to a compiled whisper graph on a thread brea… by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/58
* Add paraformer - Chinese STT by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/53
* feat:add chatTTS by @wuhongsheng in https://github.com/huggingface/speech-to-speech/pull/55
* Add ChatTTS - Chinese support  by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/59
* feat:add DeepFilterNet for speech enhancement to obtain clear speech … by @wuhongsheng in https://github.com/huggingface/speech-to-speech/pull/61
* improve documentation by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/77
* Add support for multiple languages by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/60
* Update module_arguments.py by @AgainstEntropy in https://github.com/huggingface/speech-to-speech/pull/78
* Add language arg to lightning whisper handler by @rchan26 in https://github.com/huggingface/speech-to-speech/pull/84
* Fix relative link in README by @rchan26 in https://github.com/huggingface/speech-to-speech/pull/85
* fix by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/87
* fix: Changed [True] to [False] in help text for audio_enhancement to align with actual default by @BrutalCoding in https://github.com/huggingface/speech-to-speech/pull/91
* Update: Added multi-language support for macOS. by @ybm911 in https://github.com/huggingface/speech-to-speech/pull/93
* Mac multi language by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/98
* Refactor for inference by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/106
* feat:Add rest call support similar to oepn-api style by @wuhongsheng in https://github.com/huggingface/speech-to-speech/pull/81
* Improve auto language by @eustlb in https://github.com/huggingface/speech-to-speech/pull/112
* Readme update + clarity improvements by @eustlb in https://github.com/huggingface/speech-to-speech/pull/113
* updated readme for a small typo by @ankanpy in https://github.com/huggingface/speech-to-speech/pull/115
* Fix hanging client on KeyboardInterrupt by @3manifold in https://github.com/huggingface/speech-to-speech/pull/121
* made small fixes in arguments_classes and TTS folder by @ankanpy in https://github.com/huggingface/speech-to-speech/pull/116
* Facebook mms merge by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/123
* New new faster whisper by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/124
* Add moonshine by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/127
* set keras backend to torch. by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/129
* Fixed typos in README.md by @sergiopaniego in https://github.com/huggingface/speech-to-speech/pull/137
* Bugfix: can not concatenate str + GenerationResponse by @baldassarreFe in https://github.com/huggingface/speech-to-speech/pull/144
* Update Parler-TTS base model and description by @ylacombe in https://github.com/huggingface/speech-to-speech/pull/147
* adding more languages by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/148
* multilingual improvements for parler by @andimarafioti in https://github.com/huggingface/speech-to-speech/pull/149
* Improved Error Message for get_tts_handler by @Arslan-Mehmood1 in https://github.com/huggingface/speech-to-speech/pull/155

## New Contributors
* @Vaibhavs10 made their first contribution in https://github.com/huggingface/speech-to-speech/pull/2
* @AlexHayton made their first contribution in https://github.com/huggingface/speech-to-speech/pull/7
* @RodriMora made their first contribution in https://github.com/huggingface/speech-to-speech/pull/23
* @andimarafioti made their first contribution in https://github.com/huggingface/speech-to-speech/pull/29
* @codearranger made their first contribution in https://github.com/huggingface/speech-to-speech/pull/22
* @RonanKMcGovern made their first contribution in https://github.com/huggingface/speech-to-speech/pull/40
* @wuhongsheng made their first contribution in https://github.com/huggingface/speech-to-speech/pull/48
* @rchan26 made their first contribution in https://github.com/huggingface/speech-to-speech/pull/84
* @BrutalCoding made their first contribution in https://github.com/huggingface/speech-to-speech/pull/91
* @ybm911 made their first contribution in https://github.com/huggingface/speech-to-speech/pull/93
* @eustlb made their first contribution in https://github.com/huggingface/speech-to-speech/pull/112
* @ankanpy made their first contribution in https://github.com/huggingface/speech-to-speech/pull/115
* @3manifold made their first contribution in https://github.com/huggingface/speech-to-speech/pull/121
* @sergiopaniego made their first contribution in https://github.com/huggingface/speech-to-speech/pull/137
* @baldassarreFe made their first contribution in https://github.com/huggingface/speech-to-speech/pull/144
* @ylacombe made their first contribution in https://github.com/huggingface/speech-to-speech/pull/147
* @Arslan-Mehmood1 made their first contribution in https://github.com/huggingface/speech-to-speech/pull/155

**Full Changelog**: https://github.com/huggingface/speech-to-speech/commits/2025