Jack Dorsey’s Block has released a new voice model from its research lab designed to run locally and mimic human speech with realistic expressions like laughter, pauses, and crying. The company stated that this model, named Dessa, is fully trained on human speech and requires a single NVIDIA GPU.
Dessa is a full-duplex model, meaning it can listen and speak simultaneously, similar to the GPT-Live model released by OpenAI earlier this year. While typical speech models are paired with large language models for knowledge retrieval, Block explained that Dessa’s knowledge base comes from conversational data.
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built, the company said. Talking to it feels like conversing with a person, full of emotion, nuance, pauses, and natural cadence.
Block explained that current voice AI models often sound unnatural because they force text through a pipeline to generate speech. Many tools, such as ElevenLabs v4 and Fish Audio Drama 3, use voice tags to drive emotion, whereas Dorsey's company claims its model builds emotion directly into its architecture.
Interacting with current voice AI can just plain suck. It turns your voice into text and throws away an incredible amount of information text alone can't carry. The most powerful AIs in the world can't hear how you feel. We think that's ridiculous, leads to a worse experience, and is potentially dangerous, the company stated in a blog post.
Block researchers demonstrated multiple examples of conversations with Dessa, showcasing different voices and emotions. In one instance, the bot expressed being tired from work and sounded accordingly.
We made the first AI model you can have a realistic conversation with that was trained entirely on speech! Dessa is a full-duplex speech-to-speech model that you can run on consumer GPUs. This is a GPT3 moment for speech, the team posted.
Another example showed the bot comforting a user, sounding emotional and almost breaking down.
A third example featured a Block team member playing a guessing game with the voice assistant, which guessed the answers quickly.
The company indicated that the model can operate locally on consumer-grade computers like a MacBook M4 Max or better. Currently, there is no public access to Dessa, and Block is calling for safety researchers to test the model for potential deception scenarios.
Block operates in the finance sector and sells solutions to small businesses. While Square already uses voice AI for ordering, having its own on-device voice model allows the company to power more business and consumer experiences at a lower cost.
The Editorial Staff at AIChief is a team of professional content writers with extensive experience in AI and marketing. Founded in 2025, AIChief has quickly grown into the largest free AI resource hub in the industry.