v3.85 Changelog

• Optimized UI layout by hiding infrequently used parameters by default, avoiding a cluttered interface with too many options upon startup.
• Added Speaker Recognition feature.
• Replaced the denoising model with a smaller model.
• Switched source code deployment method to uv.
• When batch translating videos, you can now specify "Sequential Serial" mode in the advanced options.
• Changed "Equal Segmentation" mode to batch inference. When using faster-whisper, batch inference can accelerate execution.
• EdgeTTS concurrency and failure retry attempts can now be individually configured in the advanced options.
• Optimized alignment strategy.
• Fixed other known issues.