Sesame
Human-like voice companions (Maya, Miles) with conversational speech models.
About
Sesame builds voice companions around its Conversational Speech Model line: Maya and Miles entered research preview in early 2025 and reached general availability in 2026 with persistent memory, scheduled workflows, web research, and per-agent context on iOS and Android. The demo-era virality came from voices that carry breath, pauses, timing, and emotional tenor — treating voice as context-aware dialogue rather than plain text-to-speech.
The research base is open: CSM-1B was open-sourced under Apache 2.0 in March 2025 and is one of the most-starred open speech releases. Sesame raised $250M in October 2025 at a $1B+ valuation and is developing lightweight AI eyewear targeted for 2027. Caveat: the shipped product voices and open base model are different artifacts — the base model has limited non-English ability and usage terms prohibit impersonation and deception.
What it can do
- Maya, Miles, Simone, and Charlie voice agents with distinct personas
- Persistent per-agent memory and incognito ephemeral mode
- Scheduled workflows and background task execution
- Open-source CSM-1B conversational speech model (Apache 2.0)
- Context-aware prosody from text plus conversation audio history
Setup
Sesame app on iOS and Android; the open CSM-1B model needs a CUDA GPU, Python 3.10, and the Transformers library.
Prerequisites
A smartphone for the companion app; a CUDA-capable GPU to self-host the open model.
Use cases
Works with
Trust score — 9/10
Scored 9/10 (verified tier, metrics-v1). Evidence: Verified via vendor site/docs/GitHub and reputable coverage, wave 79; Trending score 88/100 from researcher assessment.
Sources
Every fact above was checked against these sources. Where a claim comes from the provider and couldn't be independently verified, the text says so.