Sesame CSM

Sesame CSM TTS

A 1B conversational speech model that captures natural dialogue timing, turn-taking, and backchannel responses.

Sign up for 5,000 character limit

Wrap your text in SSML tags for precise control:

<speak><prosody rate="slow">Slow speech</prosody></speak>

Tags the selected model understands — click to drop one into your text where it happens:

This model reads plain text, so inline tags are ignored. For tag-based emotion, switch to an expressive model like Orpheus or Bark.

Define custom pronunciations (word = pronunciation):

-12 +12
0.5x 2.0x
Free with Piper, VITS, MeloTTS
Your generated audio will appear here. Choose a model, enter text, and click Generate.
Audio Generated Successfully
0:00
Download Audio Download .srt Link expires in 24h
Free tier: personal use. Commercial license from $5/mo
Love TTS.ai? Tell your friends!

About Sesame CSM

Sesame CSM (Conversational Speech Model) is a 1-billion-parameter model from Sesame designed specifically for the rhythms of human conversation. Built on a Llama backbone paired with an audio codec, it models turn-taking timing, backchannel responses (the small acknowledgements people make while listening), emotional reactions, and overall conversational flow. The result reads less like read-aloud text and more like a real spoken exchange. It is a natural fit for AI assistants, chatbots, and conversational interfaces where the goal is speech that feels responsive and human. CSM is released under Apache 2.0, and access on TTS.ai requires a Hugging Face token at the model level.

Best for: AI assistants, chatbots, conversational AI applications

Browse all Sesame CSM voices

At a glance

Developer
Sesame
License
Apache 2.0
Tier
premium
Speed
slow
Voice cloning
No
Languages
English
Max characters
500

Sesame CSM voices

Speaker 0

English
Premium Neutral

Speaker 1

English
Premium Neutral

Sesame CSM TTS — FAQ

Conversational speech. It models the natural patterns of dialogue — turn-taking timing, backchannel responses, and emotional reactions — so generated audio sounds like a real conversation rather than synthetic narration.

It is a 1-billion-parameter model built on a Llama backbone with an audio codec for waveform generation.

AI assistants, chatbots, and other conversational applications where responsive, human-sounding speech matters more than long-form narration.
← All voices