Version 1.6.0 - Chatterbox TTS & Voice Changer

[FranckyB]

Chatterbox TTS

• New TTS Engine - Integrated Chatterbox TTS (MIT license) for voice cloning with a single reference sample
• 23 Languages - English uses the fast default model; other languages auto-switch to the Multilingual model
• Voice Clone Support - Added Chatterbox as a Voice Clone engine with exaggeration, CFG weight, temperature, repetition penalty, and top-p controls
• Conversation Support - Chatterbox available as a Conversation engine with up to 8 speakers, sequential per-line generation, and configurable pauses

Voice Changer

• New Voice Changer Tool - Convert any audio to match a target voice using Chatterbox voice conversion
• Target Voice Selector - Browse and preview voice samples from your samples folder
• Source Audio Input - Upload audio files or record directly from microphone
• Save to Samples - Save converted audio directly to your samples folder for use with other tools

Conversation Tool Improvements

• Shared Voice Sample Dropdowns - Consolidated duplicate per-engine dropdowns into a single shared set of 8, reducing UI bloat
• Smart Speaker Limit - Rows 5-8 automatically hidden when using VibeVoice (4 speaker limit)

Chatterbox Multilingual Fixes

• Truncation Fix - Increased max generation tokens from 1000 to 2048 (~80 seconds max vs ~40 seconds)
• Premature EOS Fix - Fixed AlignmentStreamAnalyzer cutting audio short by requiring minimum speech frames before allowing completion
• macOS Compatibility - Added MPS/CPU support with proper device mapping and MPS availability checks

Other Improvements

• Python 3.10-3.12 Enforcement - Setup scripts now auto-discover compatible Python versions and block 3.13+
• Flash Attention Install - Added Flash Attention 2 install option to setup-windows.bat
• Prep Audio Cache Fix - Delete and clear cache now also remove LuxTTS .pt cache files

---

This discussion was created from the release Version 1.6.0 - Chatterbox TTS & Voice Changer.