Dessa is a full-duplex speech-to-speech AI model trained from scratch on human conversation with no text data. It supports natural, emotionally rich real-time voice interactions, including interruptions, laughter, sarcasm detection, and human-like cadence, making conversations feel indistinguishable from talking to a person.
Dessa is a speech-to-speech conversation model developed by a research lab at Block. Block says it was trained from scratch on human conversation audio, without text or pretrained models, and can listen and speak at the same time. Its demonstrations show conversational behaviors such as interruptions, backchanneling, laughter, and changes in vocal expression. Block says it runs locally on a MacBook Pro M4 Max or better. The page presents research questions about the model’s language learning and capabilities, alongside recordings of conversations.
