cosyvoice3

开箱即用的本地私有化部署语音服务，快速搭建Qwen3ASR/FunASR与Qwen3TTS/CosyVoice后端

CosyVoice inference in C/C++

LightTTS is a lightweight TTS inference framework optimized for CosyVoice2 and CosyVoice3, enabling fast and scalable speech synthesis in Python and supports stream and bistream modes.

To convert CosyVoice model to ONNX

CosyVoice3 text-to-speech for Unity using ONNX inference. Supports zero-shot voice cloning

Python bindings for CosyVoice3 TTS using Candle. Has the characteristics of small size, fast speed, and does not rely on libraries such as Pytorch.

CosyVoice3 LoRA fine-tuning companion repo — PEFT integration, 9 known pitfalls from IMDA NSC production runs, corrected from failed full-SFT attempt

CosyVoice3 Russian TTS on AX650N NPU (CM3588) with voice cloning and post-processing

SImple Use with CosyVoice3 and ComFyui

这是一个简单的基于cosyvoice3搭建的针对音色克隆服务端, 用于支持如本地小型AI游戏的构建和应用

Local-first AI audio production studio for stories, podcasts, immutable A/B Takes, sound design, mastering, and release packages.

Fine-tuning CosyVoice3 on Bengali podcast audio with emotion and persona instruction control. Covers the full pipeline: data collection, Gemini enrichment, ASR verification, quality filtering, and Gradio inference UI.

CosyVoice UI

FastAPI wrapper for Fun-CosyVoice3-0.5B: zero-shot voice cloning TTS with HuggingFace upload

A text-to-speech service built with CosyVoice2, with multi-node concurrency capabilities.