deerwork-ai
Featured
deer-workflow
An open-source Dynamic Workflow runtime that combines deterministic TypeScript orchestration with replaceable Agent runtimes.
CosyVoice3Pro
CosyVoice3Pro is a production-ready voice cloning service built on NVIDIA Triton Inference Server and TensorRT-LLM, featuring a decoupled speaker registry that lets you register a voice once and then generate speech with just a speaker ID and text.
Stars
19
7-day growth
No data
Forks
0
Open issues
0
License
Apache-2.0
Last updated
2026-07-29
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It provides a complete, high-performance serving stack for voice cloning with real-time benchmarks (e.g., 0.148 RTF on A100), a developer-friendly HTTP API, a web management console, and audio post-processing, all while decoupling audio feature extraction from inference to reduce per-request overhead.
Who it is for
- Developers integrating voice cloning into applications
- Teams operating high-performance TTS systems in production
- Researchers or engineers experimenting with CosyVoice-based deployment
- System administrators managing GPU-inference services
Use cases
- Build a personalized voice assistant with consistent speaker identity across sessions
- Create an API for generating custom audio content (e.g., audiobooks, voiceovers) with low latency
- Set up a scalable TTS pipeline that supports multiple speakers and on-the-fly style overrides
- Replace repeated audio uploads with a persistent speaker registry for cost savings
Strengths
- Decoupled speaker registration: register once with audio, then use speakerId+text for all subsequent TTS calls
- Proven performance: benchmarks show sub-0.6 end-to-end RTF even under 24 concurrent requests on A100-80GB
- Feature-rich: built-in speaker management (CRUD), audio post-processing (speed, volume, format), and web UI
- Developer-friendly: standard HTTP/JSON APIs (no Triton tensor protocol), plus compatibility with legacy endpoints
Considerations
- Requires NVIDIA GPU with CUDA support and Docker; no CPU fallback
- No built-in authentication for the web admin or APIs; production deployments must add external security
- Based on upstream CosyVoice (community project), not an official FunAudioLLM release
README quick start
快速开始
Description
Production-ready CosyVoice serving with Triton, reusable Speaker Registry, Public HTTP API and Web console.
Related repositories
Similar projects matched by category, topics, and programming language.
slvDev
Featured
esp32-ai
A 28.9 million parameter language model runs on an $8 ESP32-S3 microcontroller entirely on-device, generating simple stories at about 9.5 tokens per second.
jamesob
Featured
local-llm
A comprehensive guide for building and configuring a high-end local machine to run state-of-the-art LLMs, with detailed hardware choices, BIOS tuning, and Docker-based model serving.