DEV.co
AI Frameworks · 2noise

ChatTTS

ChatTTS is a generative speech synthesis model optimized for dialogue and conversational AI applications. It supports English and Chinese with fine-grained control over prosodic features like laughter, pauses, and intonation, delivering natural-sounding speech from text.

Source: GitHub — github.com/2noise/ChatTTS
39.6k
GitHub stars
4.2k
Forks
Python
Primary language
AGPL-3.0
License (OSI-approved)

Key facts

Objective fields from the source. Values we can't verify are shown as “Unknown” rather than guessed.

FieldValue
Repository2noise/ChatTTS
Owner2noise
Primary languagePython
LicenseAGPL-3.0 — OSI-approved
Stars39.6k
Forks4.2k
Open issues63
Latest releasev0.2.5 (2026-04-10)
Last updated2026-04-10
Sourcehttps://github.com/2noise/ChatTTS

What ChatTTS is

A PyTorch-based TTS model trained on 100,000+ hours of multilingual audio data, with a publicly available 40,000-hour pre-trained variant. Provides token-level and word-level prosodic control through special markup, supports multiple speaker embeddings via Gaussian sampling, and includes streaming audio generation capabilities.

Quickstart

Get the ChatTTS source

Clone the repository and explore it locally.

terminalbash
git clone https://github.com/2noise/ChatTTS.gitcd ChatTTS# follow the project's README for install & configuration

Need it deployed, integrated, or customized instead? DEV.co ships production installs.

Best use cases

LLM Assistant & Chatbot Audio Output

Designed specifically for dialogue scenarios; pairs well with language models to generate natural conversational speech for chatbot and AI assistant applications.

Interactive Voice Applications

Multi-speaker support and fine-grained prosody control enable building interactive voice experiences with varied intonation, laughter, and natural pauses.

Research & Academic Prototyping

Open-source code and model weights support TTS research, model fine-tuning, and academic exploration of dialogue-focused speech synthesis.

Implementation considerations

  • AGPL-3.0 code license is copyleft; any derivative work or integration requires source disclosure and same license, or formal licensing exception. Not suitable for closed-source commercial products without review.
  • Model weights (CC BY-NC 4.0) are explicitly non-commercial only; for-profit use is prohibited without separate commercial licensing agreement.
  • Pre-trained model is intentionally degraded (MP3 compression, high-frequency noise) to deter misuse; audio quality expectations should be set accordingly.
  • Python/PyTorch stack; GPU recommended for inference speed. Installation includes optional vLLM, FlashAttention-2, and TransformerEngine, but latest versions of these are noted as unrecommended/under development.
  • Active maintenance as of April 2026; v0.2.5 released recently. 39k+ stars suggest community interest, but formal SLAs and support channels are not standard.

When to avoid it — and what to weigh

  • Commercial Production Without Legal Review — AGPL-3.0 code license requires source code disclosure; model is CC BY-NC 4.0 (non-commercial only). Commercial use requires formal licensing review—contact [email protected].
  • Non-English/Chinese Language Support Required — Currently supports English and Chinese only. Roadmap indicates more languages are coming, but they are not yet available.
  • Production Quality Audio Without Optimization — Open-source 40k-hour model is pre-trained only (no supervised fine-tuning) and intentionally compressed with MP3 and high-frequency noise to limit misuse. Expect lower quality than internal 100k-hour version.
  • Systems Requiring Guaranteed Output Quality & Reliability — Project is ~2 years old; model reliability assertions and SLAs are not provided. Authors explicitly state they do not guarantee accuracy or completeness.

License & commercial use

Dual licensing: Code is AGPL-3.0 (strong copyleft, source disclosure required); pre-trained model weights are CC BY-NC 4.0 (non-commercial use only). Commercial use requires formal licensing review and likely separate agreement.

Code (AGPL-3.0) is not suitable for closed-source commercial products without source disclosure or licensing exception. Model weights (CC BY-NC 4.0) explicitly prohibit commercial use. For-profit deployment is not permitted under current licenses. Contact [email protected] for formal commercial licensing inquiries. Requires legal review before any commercial production use.

DEV.co evaluation signals

Editorial assessment — not user reviews. Directional, with an explicit confidence level.

SignalAssessment
MaintenanceActive
DocumentationAdequate
License clarityNeeds review
Deployment complexityModerate
DEV.co fitGood
Assessment confidenceHigh
Security considerations

No formal security audit results published. Authors intentionally added high-frequency noise and MP3 compression to discourage misuse for unauthorized voice synthesis. Plan to open-source a detection model for identifying ChatTTS-generated audio. No documented vulnerability disclosure policy or security contact. Use of pre-trained weights from Hugging Face introduces supply-chain dependency. Recommendation: conduct code review and model validation before production deployment; review AGPL-3.0 copyleft obligations if integrating into proprietary systems.

Alternatives to consider

Coqui TTS (tts library)

Open-source, permissive Apache 2.0 license, commercial-friendly. Supports multiple languages. Lower prosody control granularity; less optimized for dialogue scenarios but more deployment-flexible.

Google Cloud Text-to-Speech / Microsoft Azure Speech Services

Proprietary SaaS; commercial licenses clear and well-defined. Enterprise SLA, multiple languages, production-grade reliability. Higher cost and latency; no on-premise option.

Bark (Suno)

Open-source, MIT license, commercial-friendly. Supports multiple languages and fine-grained control. Lower maturity and community adoption; different training data and prosody model; less dialogue-optimized.

Software development agency

Build on ChatTTS with DEV.co software developers

Explore ChatTTS for conversational AI, but verify license compliance for your use case. For commercial deployment, contact the 2noise team to discuss licensing terms.

Talk to DEV.co

Related open-source tools

Surfaced by semantic similarity across the DEV.co open-source index.

Related on DEV.co

Explore the category and the services that help you build with it.

ChatTTS FAQ

Can we use ChatTTS in a commercial product?
Not without formal licensing review. Code is AGPL-3.0 (requires source disclosure for derivatives); model is CC BY-NC 4.0 (non-commercial only). Contact [email protected] for commercial licensing terms. Current licenses prohibit for-profit use.
What languages does ChatTTS support?
English and Chinese. Roadmap indicates more languages are coming, but they are not yet released. Mixed language input is supported within these two.
Is the open-source model the same quality as the internal one?
No. The public 40,000-hour model is pre-trained only (no supervised fine-tuning) and intentionally degraded with MP3 compression and high-frequency noise. The internal 100,000-hour model is higher quality and used for commercial products by 2noise.
Do you offer support or SLAs?
No formal support SLA is published. GitHub issues/PRs are welcomed. For formal inquiries, contact [email protected]. Community channels include Discord and QQ groups.

Custom software development services

DEV.co helps companies turn open-source tools like ChatTTS into production software. Our software development services cover the full lifecycle — architecture, web development, integration, and maintenance — delivered by software developers and web developers who ship. Engage our software development agency to implement or customize it for your ai frameworks stack.

Ready to Add Natural Speech to Your Dialogue Application?

Explore ChatTTS for conversational AI, but verify license compliance for your use case. For commercial deployment, contact the 2noise team to discuss licensing terms.