v4.09

kubernetes-sigs/agent-sandboxv4.09Aug 7, 2026by jianchang512

AI Summary

This release focuses on enhancing speech recognition capabilities by introducing new models like Moss-Transcribe-Diarize and Granite-Speech, while fixing bugs related to Whisper and translation prompts. It also adds support for Vietnamese and Turkish in F5-TTS.

Key Highlights

Adds Moss-Transcribe-Diarize model for speaker diarization support up to 90 minutes
Expands F5-TTS support to Vietnamese and Turkish languages
Introduces new ASR models: ARK-ASR-0.6B and granite-speech-4.1-2b
Adds Higgs Audio v3 TTS model with support for 10G+ VRAM
Improves subtitle translation prompts and fixes Whisper errors

New Features

Moss-Transcribe-Diarize model for speaker identification
ARK-ASR-0.6B and granite-speech-4.1-2b models for speech recognition
Higgs Audio v3 TTS model for text-to-speech
F5-TTS support for Vietnamese and Turkish languages
Automatic fallback to local M2M100 model when Microsoft translation fails

Full Release Notes

## v4.09 Change log

> MacOX / Linux 请使用源码部署

- 修复 faster_whisper_xxl.exe 报错
- 优化字幕翻译提示词
- 恢复导入本地源语言字幕按钮，仅单视频模式下生效
- 微软翻译已无法使用，报错时自动使用本地模型`M2M100(内置)`替换
- 语音识别 增加 [Moss-Transcribe-Diarize](https://github.com/OpenMOSS/MOSS-Transcribe-Diarize/) 模型，90分钟内音视频支持直接返回说话人
- 语音识别 Huggingface_ASR渠道 增加模型 [Audio8/ARK-ASR-0.6B](https://huggingface.co/Audio8/ARK-ASR-0.6B)  [ibm-granite/granite-speech-4.1-2b](https://huggingface.co/ibm-granite/granite-speech-4.1-2b)
- 配音渠道增加 [higgs-audio-v3-tts-4b](https://huggingface.co/bosonai/higgs-tts-3-4b) 模型，CUDA加速需10G+显存，CPU运行需20G+内存
- F5-TTS 增加 越南语、土耳其语支持
- 修复其他已知bug



----

### v4.09 (2.7G)  完整包 / Full Package for Win10/11
- 百度网盘: https://pan.baidu.com/s/1zsIDwqk8mJOLB8UUgRgLbg?pwd=1234
- huggingface.co:  https://huggingface.co/mortimerme/repocollect/resolve/main/win-pyvideotrans-v4.09.7z?download=true


### v4.09 (300MB)  补丁包 / Patch package

> 补丁包不可单独使用，仅用于覆盖旧包, 覆盖后如果闪退或，请下载完整包。

- 百度网盘: https://pan.baidu.com/s/1vTfsnVqRiRpT8ueaTvtQDg?pwd=1234
- huggingface.co: https://huggingface.co/mortimerme/repocollect/resolve/main/win-PatchUpdate-v4.09.7z?download=true

**Full Changelog**: https://github.com/jianchang512/pyvideotrans/compare/v4.08...v4.09