Block Unveils Dessa, a Human‑Like Full‑Duplex Voice Model

Local, Emotion‑Rich Speech Model Runs on Consumer Hardware

Jack Dorsey’s Block has released Dessa, a new voice model developed in the company’s research lab. The system runs entirely on a single Nvidia GPU rack and processes speech locally, delivering expressions such as laughs, pauses, and cries. Unlike many speech generators that rely on text pipelines or separate voice tags to convey emotion, Block claims Dessa’s emotional nuance is built directly into the model.

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. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence. https://t.co/r0qFLdvd6c

— jack (@jack) October 23, 2026

Dessa is a full‑duplex speech‑to‑speech model, meaning it can both speak and listen simultaneously, similar to OpenAI’s GPT‑Live. The knowledge base comes from conversational data rather than a separate large language model, allowing the system to retain the subtleties of human dialogue.

In a blog post, Block criticized existing voice AI for discarding emotional cues: “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.”

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. pic.twitter.com/XVD8sSyJn8

— Ragavan Thurairatnam (@RagavanThurair) October 23, 2026

The company shared several demonstration videos. In one, the assistant mentions feeling tired and its voice noticeably slows and softens. Another clip shows the model expressing comfort while sounding on the verge of tears. A third example features a Block team member playing a guessing game with the assistant, which answers correctly and quickly.

Dessa=full‑duplex speech model vs Human=me, https://t.co/1eI6QFHo1w pic.twitter.com/zlZfxH2BA5

— Aydin Polat (@AydinPolatDessa) October 23, 2026

For the love of the game.Watch Cara play a guessing game with Dessa. The model guesses the first one right straight away, and the second doesn’t take long either.https://t.co/FDmoXNPpIh pic.twitter.com/I9f8saduEA

— Block (@blocks) October 23, 2026

Availability and Safety Considerations

Dessa can operate on consumer‑grade computers such as a MacBook M4 Max or higher, enabling on‑device use without cloud connectivity. The model is not yet publicly available; Block has invited safety researchers to test it for scenarios where a voice agent might deceive or manipulate users.

Business Impact

Block, primarily known for its financial services and Square point‑of‑sale solutions, already employs voice AI for ordering. With an on‑device, human‑like voice model, the company anticipates expanding both business and consumer applications while reducing reliance on external services and lowering costs.

By embedding emotion, nuance, and natural cadence directly into the model, Block positions Dessa as a potential milestone for conversational AI, akin to the breakthrough that GPT‑3 represented for text generation.