RepositoryQwenAudio/CosyVoiceTool
Engineers building conversational voice agents or media pipelines gain access to a production-ready, low-latency TTS model that runs entirely on self-hosted infrastructure without cloud lock-in.
CosyVoice by QwenAudio is a full-stack open-source framework covering voice generation model training, inference, and deployment. Relying on large language models, the repository provides speech synthesis with zero-shot voice cloning across diverse languages.
Fun-CosyVoice 3.0 removes the need for standalone text-normalisation frontends, directly handling raw formats, numbers, and symbols. In addition, the system enables fine-grained instruction control, letting callers configure dialects, emotions, volume, and speech tempo. To rectify mispronunciations in production, CosyVoice supports targeted pronunciation inpainting via English CMU phonemes and Chinese Pinyin.
For conversational interfaces, CosyVoice implements bi-streaming (streaming text input paired with streaming audio output). It achieves turnaround latencies as low as 150 ms. Runtime integrations with vLLM and NVIDIA Triton TRT-LLM, combined with GRPO reinforcement learning training scripts, prepare the model for high-throughput deployment.
Backed by tens of thousands of GitHub stars, comprehensive benchmarks, and an Apache 2.0 licence, CosyVoice presents a viable alternative to proprietary voice generation APIs for self-hosted AI applications.
Summary generated by AI (Gemini 3.8 Flash, 1 November 2026). It may contain errors – the original source is authoritative.
Similar in content, found by the AI search.
Repository:huggingface/speech-to-speech
Hugging Face has released speech-to-speech, a modular open-source framework designed to build voice agents with open-source models. The pipeline spans VAD, STT, LLM, and TTS, exposing the core OpenAI Realtime GA event set over WebSocket and WebRTC.
13kstarsPython
Repository:jamiepine/voicebox
Voicebox is a local-first, open-source voice studio combining text-to-speech, voice cloning, and system-wide dictation. It serves as a self-hosted alternative to proprietary cloud solutions such as ElevenLabs and WisprFlow.
55.6kstarsTypeScript
X post:MIKE
An X post highlights an open-source project by a solo developer that runs text-to-speech and voice cloning entirely on local hardware. The repository has accumulated 19.4k GitHub stars, presenting itself as a free self-hosted alternative to ElevenLabs.
21.1kbookmarks1.1mviews
Repository:m-bain/whisperX
WhisperX extends OpenAI's Whisper model with VAD-based batching, word-level alignment using wav2vec2, and speaker diarization via pyannote-audio. It is aimed at developers seeking precise, high-throughput audio transcription running up to 70x real-time.
25kstarsPython
Repository:ollama/ollama
Ollama is an open-source Go framework designed to run open-weight models locally, manage them via APIs, and connect them directly to coding tools and conversational assistants.
180.2kstarsGo
Repository:calesthio/OpenMontage
OpenMontage is an open-source video production framework designed to turn AI coding assistants such as Cursor or Claude into automated video production studios. Built with Python, it provides 12 production pipelines, over 100 tools, and 700+ skill and knowledge files connecting speech, image, and video generation services with programmatic editing engines.
56.2kstarsPython
Parts of this content were created with the assistance of AI.
