Block has released a new voice model called Dessa from its research lab, and the company says it runs locally on a consumer computer and can convey emotions such as laughter, pauses and crying. Block, a company that focuses on finance and sells tools to small businesses, announced the model in posts dated October 23, 2026. The company said Dessa was trained entirely on human speech using a single NVIDIA graphics processor.
How Dessa works
Dessa is a full-duplex model, a term for a system that can talk and listen at the same time. Block compared this ability with GPT-Live, a model OpenAI released earlier this year. Speech systems are usually paired with a large language model, the type of AI that answers questions using stored knowledge. Block said Dessa's knowledge comes from the conversational data it was built on.
The company argues that most voice AI converts speech into text first, which discards tone and feeling before the system responds. Block said its model does not work that way and that emotion is built into it. Several rival products, including ElevenLabs v4 and Fish Audio Drama 3, use voice tags to steer emotion in generated speech. Block has not made clear whether Dessa uses any similar tags.
Examples and reactions
Block's researchers posted several sample conversations in which the assistant used different voices, speaking speeds and emotions. In one, the bot said it was tired from work, and its voice then sounded tired. In another, the bot spoke about the person who was making it comfortable and sounded close to breaking down. A third clip showed a Block team member playing a game with the voice assistant.
In a blog post, the company criticised current voice AI, saying it discards a great deal of information that text cannot carry. It also said that AI systems that cannot hear how a person feels lead to a worse experience and could be dangerous. These are Block's own views, and independent testing of them has not been reported.
Access and what comes next
Block said Dessa can run on a consumer-grade computer such as a MacBook M4 Max or better, so it does not need a large data-centre system to operate. The model is not publicly accessible at the moment. Block has put out a call for safety researchers to test it, with a focus on cases where a voice agent could trick people on the other end.
Block already uses voice AI in Square, which handles ordering for some businesses. The company said an on-device voice model could support more business and consumer uses at a lower cost. Block has not said when Dessa will be released to the public.
Aasan alfaaz mein
Block ne Dessa naam ka naya voice model banaya hai jo locally, yani apne computer par, chal sakta hai. Yeh model baat karte hue sun bhi sakta hai aur hansi, wiqfe aur jazbaat ka andaaz bhi dikha sakta hai. Abhi yeh aam logon ke liye available nahi hai, aur company ne safety researchers ko iska test karne ke liye bulaya hai. Aapke liye iska matlab hai ke aane wale waqt mein voice assistants zyada natural lag sakte hain.
