AI company Sesame has released CSM-1B, the base model behind its advanced voice assistant, Maya. The 1-billion parameter model, now available under an Apache 2.0 license, allows for commercial use with minimal restrictions.
CSM-1B generates RVQ audio codes from text and audio inputs using residual vector quantization (RVQ), a technique also utilized in AI audio technologies from Google and Meta. The model combines a Meta Llama-based architecture with a custom audio decoder, enabling it to produce a variety of voices.
Sesame, co-founded by Brendan Iribe, has raised funding from Andreessen Horowitz, Spark Capital, and Matrix Partners. Beyond AI assistants, the company is developing AI-powered smart glasses designed for all-day wear.
