text-to-speech / voice generation
CosyVoice is a multi-lingual, LLM-based text-to-speech and voice-cloning model that provides inference, training, and deployment code.
Same problem, same approach. Swapping one for another is a config change, not a rewrite.
Choose when you want a current Apache-licensed multilingual TTS with streaming and free-form voice design from the Qwen ecosystem.
Qwen3-TTS is an open-source series of TTS models developed by the Qwen team at Alibaba Cloud, supporting stable, expressive, and streaming speech generation, free-form voice design, and vivid voice cloning.
Choose when you need voice cloning for more than 600 languages with an Apache-2.0 license.
High-Quality Voice Cloning TTS for 600+ Languages
Choose when you prefer a compact open-weight multilingual TTS model with strong expressiveness and permissive licensing.
Zonos-v0.1 is a leading open-weight text-to-speech model trained on more than 200k hours of varied multilingual speech, delivering expressiveness and quality on par with—or even surpassing—top TTS providers.
Choose when ultra-fast inference (150x realtime) for voice-cloning TTS outweighs CosyVoice's dialect coverage.
A high-quality rapid TTS voice cloning model that reaches speeds of 150x realtime.
Choose when you want a recent multilingual TTS family from the OpenMOSS community with high-fidelity speech and sound generation.
An open-source model family for long-form speech, dialogue synthesis, voice design, sound effects, and real-time streaming TTS
Choose when you have very little voice data (about 1 minute) and need rapid few-shot voice cloning under an MIT license.
1 min voice data can also be used to train a good TTS model! (few shot voice cloning)
Choose when you need instant voice-color cloning with cross-lingual support and a lightweight MIT-licensed model.
Instant voice cloning by MIT and MyShell. Audio foundation model.
Choose when you want a comparable zero-shot TTS with its own inference pipeline and model weights from the Spark project.
Spark-TTS Inference Code
Choose when you need a multilingual/cross-lingual zero-shot TTS engine from Youdao with permissive open source.
Confucius4-TTS: a Multilingual and Cross-Lingual Zero-Shot TTS Engine
Choose when you need a production-oriented zero-shot TTS with controllable output and efficient deployment.
An Industrial-Level Controllable and Efficient Zero-Shot Text-To-Speech System
Choose when you need zero-shot TTS plus speech editing in the wild, with an LLM-based approach similar to CosyVoice.
Zero-Shot Speech Editing and Text-to-Speech in the Wild
Solves the same problem with a different architecture or at a different layer. Expect to rewrite the integration.
Choose when you want a TTS built by inverting Whisper, offering a different prosody and architecture from LLM-based CosyVoice.
An Open Source text-to-speech system built by inverting Whisper.
Choose when you prefer a tokenizer-free TTS architecture rather than the LLM-token approach of CosyVoice.
VoxCPM2: Tokenizer-Free TTS for Multilingual Speech Generation, Creative Voice Design, and True-to-Life Cloning
Choose when you need controllable TTS covering 7000+ languages and fine-grained prosody control, not zero-shot voice cloning.
Controllable and fast Text-to-Speech for over 7000 languages!
Choose when your primary need is editing emotion/style in existing audio, with TTS also available.
A powerful 3B-parameter, LLM-based Reinforcement Learning audio edit model excels at editing emotion, speaking style, and paralinguistics, and features robust zero-shot text-to-speech
An earlier or more limited way to do the job. Still the right call when you need something small, proven, or CPU-only.
An older expressive TTS model; choose for its prosody style if you accept no recent maintenance.
Foundational model for human-like, expressive TTS
A lighter multilingual TTS without zero-shot voice cloning; choose for low-resource or highly efficient deployment.
High-quality multi-lingual text-to-speech library by MyShell.ai. Support English, Spanish, French, Chinese, Japanese and Korean.
An older Tacotron-2-based cross-lingual voice cloning codebase; choose only for legacy research or education.
Tacotron 2 - PyTorch implementation with faster-than-realtime inference modified to enable cross lingual voice cloning.
Does the same job, but ships as an app. Useful to a person, not swappable into a codebase.
A Gradio WebUI that wraps multiple TTS/voice-cloning engines including CosyVoice; choose for GUI experimentation, not code integration.
Gradio WebUI for creators and developers, featuring key TTS (Edge-TTS, kokoro) and zero-shot Voice Cloning (E2 & F5-TTS, CosyVoice), with Whisper audio processing, YouTube download, Demucs vocal isolation, and multilingual translation.
Notebooks for voice cloning with Bark, an end-user recipe rather than a library to integrate.
Bark Voice Cloning and Voice Cloning for Chinese Speech
A ComfyUI custom node integration that wraps various TTS engines including CosyVoice; use as a GUI workflow tool, not a model dependency.
A ComfyUI custom node integration for local multi-engine multi-language Text-to-Speech and Voice Conversion. Supports: RVC, Echo-TTS, Qwen3-TTS, Cozy Voice 3, Step Audio EditX, IndexTTS-2, Chatterbox (classic and multilingual), F5-TTS, Higgs Audio 2, 3, and VibeVoice with unlimited text length, SRT timing, Character support, and many audio tools
These projects were analysed and named CosyVoice among their alternatives. The relationship is not symmetric — how CosyVoice rates them is a separate judgement, made when CosyVoice is analysed in its own right.
calls CosyVoice “Choose it for Apache-2.0 licensing, full-stack training/deployment, and broad multilingual support.”
1 min voice data can also be used to train a good TTS model! (few shot voice cloning)
calls CosyVoice “choose when you need a full-stack solution including training and deployment, not just inference.”
Instant voice cloning by MIT and MyShell. Audio foundation model.
calls CosyVoice “Choose for a full-stack multilingual TTS with training and deployment support.”
SoTA open-source TTS
calls CosyVoice “choose for Apache-2.0 full-stack multilingual TTS with training support.”
An Industrial-Level Controllable and Efficient Zero-Shot Text-To-Speech System
calls CosyVoice “Choose when you need a full-stack multilingual voice generation toolkit with training and deployment.”
Qwen3-TTS is an open-source series of TTS models developed by the Qwen team at Alibaba Cloud, supporting stable, expressive, and streaming speech generation, free-form voice design, and vivid voice cloning.
calls CosyVoice “Use for full-stack training/inference/deployment of multilingual TTS.”
High-Quality Voice Cloning TTS for 600+ Languages
calls CosyVoice “choose over Zonos for a production-tested TTS stack with training/deployment utilities and strong Chinese/English support from the Qwen team.”
Zonos-v0.1 is a leading open-weight text-to-speech model trained on more than 200k hours of varied multilingual speech, delivering expressiveness and quality on par with—or even surpassing—top TTS providers.
calls CosyVoice “Choose when you want a full-stack multilingual TTS with training/deployment support and a large ecosystem.”
VoxCPM2: Tokenizer-Free TTS for Multilingual Speech Generation, Creative Voice Design, and True-to-Life Cloning
calls CosyVoice “For a large multilingual voice generation model with full-stack training, if you need that scale.”
Lightning-Fast, On-Device, Multilingual TTS — running natively via ONNX.
calls CosyVoice “Choose when you need large-scale multilingual voice generation with full training/deployment stack; local and heavy.”
Use Microsoft Edge's online text-to-speech service from Python WITHOUT needing Microsoft Edge or Windows or an API key
calls CosyVoice “choose for full-stack training/inference support and multi-lingual voice generation.”
Spark-TTS Inference Code
calls CosyVoice “choose if you need a full-stack multilingual TTS with training support and GPU deployment.”
A TTS that fits in your CPU (and pocket)
calls CosyVoice “choose for a full-stack multilingual TTS with training and deployment support, no speech editing.”
Zero-Shot Speech Editing and Text-to-Speech in the Wild
calls CosyVoice “choose for a complete multilingual TTS stack, but it is heavier than EmotiVoice.”
EmotiVoice 😊: a Multi-Voice and Prompt-Controlled TTS Engine
calls CosyVoice “Large multilingual voice generation model with full-stack ability; choose it for extensive language coverage and training options.”
High-quality multi-lingual text-to-speech library by MyShell.ai. Support English, Spanish, French, Chinese, Japanese and Korean.
calls CosyVoice “choose it when you need a full-stack multilingual generative TTS with training and deployment tooling.”
eSpeak NG is an open source speech synthesizer that supports more than hundred languages and accents.
calls CosyVoice “Choose for full-stack multilingual voice generation with training and deployment if you want a larger ecosystem.”
A powerful 3B-parameter, LLM-based Reinforcement Learning audio edit model excels at editing emotion, speaking style, and paralinguistics, and features robust zero-shot text-to-speech
calls CosyVoice “Choose when you need a production-ready multilingual TTS with full training/deployment and don't require MLX/Apple Silicon.”
A text-to-speech (TTS), speech-to-text (STT) and speech-to-speech (STS) library built on Apple's MLX framework, providing efficient speech analysis on Apple Silicon.