Jack Dorsey's Block Unveils Dessa, a Locally-Run Voice Model With Human-Like Emotion
Jack Dorsey's Block has released a new voice model from its research lab that runs locally and sounds strikingly human, complete with expressions such as laughs, pauses, and cries. The company said the model, called Dessa, is trained entirely on human speech and can run on a single NVIDIA GPU.
A Full-Duplex, Speech-Native Architecture
Dessa is a full-duplex model, meaning it can talk and listen at the same time, much like the GPT-Live model OpenAI released earlier in 2026. Typically, speech models are paired with a large language model for knowledge retrieval. Block said Dessa's knowledge stems directly from the conversational data it was trained on.
A Departure From Tag-Driven Emotion
Block argues that current models can sound unnatural and annoying because they rely on a text pipeline to drive speech generation. Many models, such as ElevenLabs v4 or Fish Audio Drama 3, use voice tags to steer emotion during speech generation. Dorsey's company claims that emotion is built into its model. It remains unclear whether that emotion is being steered by some kind of tags.
"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 said in a blog post.
Demonstrated Emotional Range
The company's researchers posted multiple examples of conversations with Dessa in which the agent adopted different voices, cadences, and emotions. In one conversation, the bot said it was tired from work and, at that point, sounded tired.
In another example, the bot spoke about the person making it comfortable while sounding emotional and almost breaking down. A further demonstration showed a Block team member playing a guessing game with the voice assistant.
Positioning in the 2026 Voice AI Landscape
The release places Dessa alongside a wave of 2026 voice model launches, including OpenAI's GPT-Live for more natural live conversations and ElevenLabs v4 with expanded expression control across 90 languages. Block's researchers have framed Dessa's speech-native, full-duplex design as a significant leap. As Ragavan Thurairatnam put it on X: "This is a GPT3 moment for speech."
via TechCrunch AI