Sesame builds conversational voice AI that became notable for a specific reason: its demo voices sound so naturally human, complete with breath, hesitation, and emotional inflection, that listeners routinely fail to identify them as AI-generated.

Rather than optimizing purely for clarity or speed, Sesame's Conversational Speech Model targets what the company calls the "voice presence" problem, the subtle prosody, timing, and emotional coloring that separates a voice you enjoy talking to from one you merely tolerate. The company has released research and open components around its approach, and positions its technology for voice assistants, companions, and any application where extended natural conversation, not just isolated commands, is the actual goal.

Who is it for?

Developers building voice-first products where conversational naturalness matters more than raw throughput, and companies exploring AI companions or assistants meant for extended dialogue rather than quick queries.

How much does Sesame cost?

Access details and pricing are evolving as the company moves from research demos toward productized offerings; check the current site for API and licensing terms.

Our verdict

Sesame's demos represent a genuine step change in perceived voice naturalness, arguably ahead of the field on the specific "does this feel human to talk to" dimension. As the technology productizes further, it is one of the most closely watched voice AI companies right now.