[KongchangAI]
ModelCSM

Sesame

A series of conversational speech models, such as CSM, released by Sesame AI, a company founded by members of the former Oculus team. The series emphasizes low-latency speech generation with emotional and prosodic variation, and its demonstrations produce highly realistic speech.

Timeline (last 90 days)

Oct 27

Some users speculate that Dessa's technical lineage may be moshi -> sesame -> dessa, but this claim has not been officially confirmed.

Unverified50%
Oct 27

Sesame is a conversational voice model (such as the CSM series) released by Sesame AI, a company founded by the former Oculus team.

Unverified50%
Oct 27

Sesame drew wide attention due to a demo publicly released in early 2025, with many listeners saying its voice is already hard to distinguish from a real person.

Unverified50%
Oct 27

Sesame's core idea is to treat speech as a generative task that requires understanding context and conversational intent, rather than simple text-to-speech.

Unverified50%

All Facts (4)

Source Articles