Sesame · 2025– · conversational
Sesame’s Conversational Speech Model (CSM-1B weights on Hugging Face) is a dialogue-native speech model, closer to Moshi than to Piper. The clips are from Sesame’s 2025 research post ‘Crossing the uncanny valley of voice’ — conversation and contextual-expressivity demos.
The ‘this doesn’t sound like TTS’ conversational demo.
Vendor / paper demo
CSM multi-speaker conversation · Official Sesame CSM conversation sample from Crossing the uncanny valley of voice (WAV converted to MP3).
Vendor demo: Sesame CSM conversation clip (sesame_conversation_01) from the research post. Exact spoken text is not printed next to the player; listen to the file.
Vendor / paper demo
CSM contextual expressivity · Official Sesame CSM expressivity sample from Crossing the uncanny valley of voice (WAV converted to MP3).
Vendor demo: Sesame CSM contextual-expressivity clip (sesame_contextual_expressive_01) from the research post. Exact spoken text is not printed next to the player; listen to the file.