TechCrunch · AI· Ivan Mehta·· 26 days agoAI score60
Block releases speech model Dessa, which can run locally on consumer-grade GPUs
Block’s new voice model sounds eerily human
AI briefing
Block, the research lab under Jack Dorsey, has released the speech model Dessa. The model is trained entirely on human speech, can run locally on a single NVIDIA GPU, and can speak and listen at the same time. Block calls Dessa a full-duplex speech-to-speech model, with expressions such as emotion, pauses, and laughter built into the model, and it can run on a MacBook M4 Max or higher-spec consumer computer. Dessa is not yet publicly available, and Block is inviting security researchers to participate in testing.
Open original techcrunch.com
OpenAI / SoraDubbing and AudioModel Releases#Model release#Voice & audio#On-device & realtime#OpenAI