This AI can speak like a real human, and you can run it locally on your own computer.

I don't think you understand the gravity of this situation, here listen to this.pic.twitter.com/pILsdJonSB https://t.co/RPjdT9X6PY

— Screen (@Screeendev) October 24, 2026

Block announced Dessa on October 23, 2026. It is a full-duplex speech-to-speech model trained from scratch only on human conversation audio with no text and no pretrained models underneath. The company says inference runs faster than real time and that the model runs locally on a MacBook Pro with an M4 Max chip or better. It is currently a research preview behind a waitlist with early access limited to a small group of AI safety researchers. No public weights, API or benchmarks have been released. All available demos come from Block itself. The “almost indistinguishable from a real person” claim matches the company’s description of its own recordings but has not been independently tested.

🇺🇸 Jack Dorsey’s new AI voice model Dessa is absolutely insane.

There is no way to distinguish it from a human, man … crazy.

Thats kinda cool, but many people will get attached to it and isolate from real humans even more. Many will get scammed on the phone, especially… https://t.co/EY0343k6mN

— Lord Bebo (@MyLordBebo) October 24, 2026

Block official page — Oct 23, 2026: https://block.xyz/dessa
Voice AI Space — Oct 24, 2026: https://www.voiceaispace.com/articles/block-launches-dessa-a-full-duplex-speech-model-trained-only-on-audio
WebProNews — Oct 24, 2026: https://www.webpronews.com/block-dessa-speech-to-speech-model-no-text/
Unite.AI — Oct 23, 2026: https://www.unite.ai/block-unveils-dessa-a-full-duplex-speech-model-trained-on-conversation/