Skip to content
← Models
Open weights2Noise/10 trustMetrics-checked
ChatTTS logo

ChatTTS

Conversational TTS for Chinese and English — laughter, pauses, and dialogue dynamics from text.

Visit official page →

About

ChatTTS (May 2024) is 2Noise's text-to-speech model built specifically for conversational scenarios: dialogue with laughter, pauses, interruptions, and mixed Chinese/English. It was one of the first open models to make casual, podcast-style speech synthesis practical.

Licensing is the catch — read carefully. The GitHub code is AGPL-3.0 (copyleft: network-deployed modifications must share source), while the Hugging Face model weights are CC-BY-NC-4.0 (non-commercial). Either way, commercial products need a different model or a license from 2Noise. For research, personal projects, and AGPL-compatible open work, it remains a capable conversational TTS option.

Specs

Context window

N/A (text-to-speech model)

Pricing

Free for non-commercial use; commercial use restricted (CC-BY-NC-4.0 weights, AGPL-3.0 code).

Free quota

Free — download and run the weights yourself for non-commercial purposes.

API access

Self-hosted (Python inference code on GitHub; Hugging Face checkpoints).

Weights

Open weights

Released

May 1, 2024

Benchmarks

No standardized benchmarks. Community reception (2024): strong for its conversational niche at release; since superseded by newer models (F5-TTS, CosyVoice2, IndexTTS) on quality.

Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.

Use cases

ttspodcastingvoice-agentsaudio

Trust score — 5/10

/10 trustMetrics-checked

Scored 5/10 (metrics tier). Pioneering conversational TTS with real community adoption, but the score reflects hard limits: (1) dual restrictive licensing — AGPL-3.0 code plus CC-BY-NC-4.0 weights — rules out most commercial use; (2) the model is 2+ years old and superseded on quality by F5-TTS, CosyVoice2, and IndexTTS; (3) no active development signal from 2Noise. A historical landmark, not a current recommendation for new builds.

How the trust score is built

Full methodology →
Adoption · stars/downloads, log-scaled vs peers
5.4/10
Maintenance · recency, releases, issue load
4.0/10
Openness · license + docs
4.6/10
Manual review · sanity read + red-flag check
5.0/10
  • · HF downloads 2,417 (peer-normalized)
  • · released 895 days ago
  • · manual review: pioneering conversational TTS but 2+ years old, superseded on quality, no active development signal; license: dual restrictive: AGPL-3.0 code + CC-BY-NC-4.0 weights — no commercial use

Sources