Skip to content

v1.7.1

Compare
Choose a tag to compare
@Huanshere Huanshere released this 11 Nov 07:09
· 132 commits to main since this release

Release Notes

🚀 New Features:

  • Added MPS support for Demucs to improve performance.
  • Implemented an error retry mechanism for batch processing.
  • Set the default to use the Gemini model and updated documentation accordingly.
  • Added an auto-update feature for ytdlp.

🐛 Bug Fixes:

  • Corrected a long video segmentation error.
  • Fixed loading issues with the local Chinese Whisper model.
  • Improved audio splitting robustness and encoding handling.
  • Resolved issues with handling reference audio prerequisites for GPT-SoVITS batch processing.
  • Correctly implemented retry on translation failures.

🔧 Improvements:

  • Updated the CPU-specific torch version in the installation process.
  • Refactored to simplify the prompt reasoning chain due to minimal improvement.

📝 Updates:

  • Removed the install option in the OneKey batch script.
  • Added a section for the SaaS website in the documentation.

🚀 新功能:

  • 为 Demucs 增加了 MPS 支持以提升性能。
  • 实现了批处理的错误重试机制。
  • 设置默认使用 Gemini 模型并相应更新了文档。
  • 新增 ytdlp 自动更新功能。

🐛 问题修复:

  • 修正了长视频分段错误。
  • 修复了本地中文 Whisper 模型的加载问题。
  • 改进了音频分割的鲁棒性和编码处理。
  • 解决了 GPT-SoVITS 批处理的参考音频前提条件问题。
  • 正确实现了翻译失败时的重试机制。

🔧 改进:

  • 更新了安装过程中与 CPU 相关的 torch 版本。
  • 重构以简化提示推理链条,因其改进效果有限。

📝 更新:

  • 移除了 OneKey 批处理脚本中的安装选项。
  • 在文档中添加了 SaaS 网站的部分。