A modular, highly efficient infrastructure for fine-tuning both Chatterbox TTS (Standart) and Chatterbox Turbo models with your own dataset and generating high-quality speech synthesis. This kit is specially designed to support new languages and voices by intelligently extending the model's vocabulary. With the newly added LoRA (Low-Rank Adaptation) support, you can now train high-quality voices faster and with significantly less VRAM. This repository allows you to choose your training strategy using the islora flag in src/config.py. What is it? LoRA (Low-Rank Adaptation) freezes the massive base model and only trains tiny adapter layers alongside the new language embeddings. What is it?…