Block has unveiled Dessa, a full-duplex speech-to-speech model trained exclusively on human conversation, featuring live demos, a public waitlist, and early…
1 editorial report · 1 verified social mention. The most authoritative report leads while later evidence completes the story.
Block's research lab has introduced Dessa, a full-duplex voice AI model designed to sound remarkably human, complete with pauses, laughs, and cries. Trained solely on human speech using a single NVIDIA GPU, the model can run locally on high-end consumer hardware like an M4 Max Mac. Block notes that traditional voice AIs often sound robotic due to text-based pipelines. While public access is not yet available,… https:// en.hacks.gr/i-block-paroysiase -ena-neo-systima-technitis-noimosynis-poy-milaei-me-entyposiaka-anthropini-foni/ # ArtificialIntelligence # MachineLearning # Block # VoiceAI
Open mention