Sesame CSM-1B
Open 1B conversational speech model — generates lifelike, context-aware dialogue audio from text and audio history.
About
CSM (Conversational Speech Model) is Sesame's 1-billion-parameter open-weights speech model, released March 2025 alongside the viral Maya/Max demo voices. Unlike classic TTS that reads text aloud, CSM generates RVQ audio codes conditioned on both text and prior audio context, so it captures conversational dynamics — interruptions, backchanneling, prosody that responds to the dialogue.
Released under Apache 2.0 (commercial use allowed). The 1B size runs on a single consumer GPU. Sesame's demos set a new bar for conversational realism in 2025, and the open release let the community build on it (several fine-tunes and real-time wrappers exist). English-focused; long-form narration is not its strength — it is tuned for dialogue. Voice-cloning misuse risk applies as with all such models.
Specs
Context window
Pricing
Free quota
API access
Weights
Released
Benchmarks
No standardized benchmarks. Provider-claimed (Sesame demo comparisons): state-of-the-art conversational naturalness — treat as provider-claimed. Independent reception (Mar 2025): widespread press coverage of the demo's realism; community fine-tunes report strong dialogue quality.
Provider-reported unless a source says otherwise. Benchmarks are marketing-adjacent by nature — treat them as directional.
Use cases
Trust score — 7/10
Scored 7/10 (metrics tier). Genuinely open (Apache 2.0), solid adoption (131k HF downloads), and the conversational-audio approach is technically distinctive with broad independent coverage of the demos. Docked three points: (1) quality claims rest on Sesame's own demos, not independent benchmarks; (2) released March 2025 with no follow-up checkpoint — maintenance signal is aging; (3) it is dialogue-tuned — buyers wanting narration or single-speaker TTS should look at purpose-built models instead.
How the trust score is built
Full methodology →- Adoption · stars/downloads, log-scaled vs peers
- 7.6/10
- Maintenance · recency, releases, issue load
- 5.0/10
- Openness · license + docs
- 9.4/10
- Manual review · sanity read + red-flag check
- 8.0/10
- · HF downloads 130,944 (peer-normalized)
- · released 591 days ago
- · manual review: distinctive conversational-audio approach, broad independent press coverage of demos; quality claims rest on Sesame's own demos; license: Apache-2.0