🔍 Read the full analysis: A Human Voice—or Human-Like AI? Block’s Dessa Keeps The Question Open on ThorstenMeyerAI.com
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Block announced Dessa on October 23, describing it as a speech-to-speech research model trained on recorded human conversation. Company-selected demos show expressive, interruptible conversation, but outsiders have not tested the model and Block has released no paper, benchmarks, weights or API.
Block announced Dessa, a research model designed to listen and speak at the same time, with company demos showing laughter, interruptions and emotionally expressive responses. The announcement puts a more natural-sounding voice system in view, but no independent researcher or reporter has tested the model directly, and Block has not published a technical paper or made it publicly available.
Block, the company behind Square and Cash App, published a project page for Dessa on October 23. It says the model was trained from scratch on human conversation, without text or pretrained models, and can handle speech in a full-duplex manner: listening while talking, cutting in, and responding with backchannel cues. Block also says the system recognizes and produces features such as sarcasm, laughter and sighs.
The demonstrations on Block’s page include a model laughing mid-sentence, helping a speaker recall a film title, and responding to news about an anniversary surprise. Block says the clips are live recordings and that nobody steered the conversations. Those are company-provided examples, not independent evaluations. The Verge reported that some exchanges could be awkward or uncanny, while finding Dessa more natural-sounding than another voice model in the comparison it heard.
Block says inference runs faster than real time on a MacBook Pro M4 Max or better and that training used one NVIDIA GB200 NVL72. It has not provided a latency figure, model size, benchmark results, technical paper, model card, license or public release date. A waitlist and early access for a small group of AI safety researchers are the access options described so far.
Infographic 01 · Dessa fact sheet
What Block claims, what was heard, what is missing.
Every capability below comes from Block or its selected recordings. Nobody outside Block has run the model yet.
| Topic | What we know | Status |
|---|---|---|
| Training data | Human conversation only: “no text, no pretrained models” | Block’s claim |
| Conversation | Full-duplex: listens while it talks, backchannels, cuts in | Block’s claim |
| Emotion | Hears and speaks laughter, sarcasm, sighs; unclear if steered by tags (TechCrunch) | Block’s claim |
| Speed | “Faster than real time”; no latency figure | Block’s claim |
| Hardware | Trained on one NVIDIA GB200 NVL72; runs on a MacBook Pro M4 Max or better | Block’s claim |
| Demos | Some chats “a little awkward and uncanny”, more natural than GPT-Live-1 (The Verge) | reporters heard it |
| Size, benchmarks, release date | Not disclosed (CoinDesk) | not disclosed |
| Paper, model card, license, weights | None found as of 26 Oct 2026 | not disclosed |
| Languages | Not listed; Block calls learning new languages an open question | not disclosed |
| Voice cloning safeguards | No statement on watermarking, consent checks or imitating a specific person | not disclosed |
Who can use it
research preview
- Public: a waitlist; “launch links are coming soon”
- Early access: “a small group of AI safety researchers”, institutional email required
- Business use: none announced; Square already uses voice AI for ordering (TechCrunch)
Who built it
from Block’s page
- A “(very) small AI research lab at Block” reporting to Jack Dorsey
- Founded by former members of Dessa, the Toronto startup Square acquired in 2020 (CoinDesk)
- Lead Ragavan Thurairatnam: “a GPT3 moment for speech”
Sources: Block, Dessa project page (23 Oct 2026); Ragavan Thurairatnam on X (23 Oct 2026); The Verge (23 Oct 2026); TechCrunch (24 Oct 2026); CoinDesk (24 Oct 2026).
Powered by Thorsten Meyer AI
When Natural Voices Meet Fraud Risks
A voice that can respond fluidly, react to emotion and interrupt naturally could make voice interfaces easier to use. The same qualities may also make it harder for a person on a call to distinguish a synthetic voice from a human one. That concern does not mean Dessa has been used in fraud; the model has not been publicly released, and no evidence in the available reporting links it to a scam.
The risk matters because voice cloning is already part of reported fraud. The FBI’s 2025 Internet Crime Report, published in April 2026, counted 22,364 complaints mentioning AI and $893 million in reported losses. The report also cited more than $5 million in losses from distress scams involving cloned voices of loved ones. These figures describe reported complaints and losses across the specified period; they do not establish that Dessa or any particular model caused them.
For businesses and families, the practical lesson is not to treat a familiar-sounding voice as proof of identity. Independent verification—such as calling back using a known number and confirming unusual requests through a separate channel—can reduce reliance on voice alone. A family code word may offer another check, although it should not replace verification when money or sensitive information is involved.
Products Worth Considering
Bose QuietComfort Wireless Noise-Cancelling Over-Ear Headphones, Black
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
As an affiliate, we earn on qualifying purchases.
What Block Says It Built
Block describes Dessa as the work of a small research lab associated with its acquisition of Toronto-based deep-learning startup Dessa in 2020. The project’s premise is that speech carries information beyond the words themselves: timing, emotion and conversational cues. Block argues that conventional systems built around speech-to-text can discard some of those signals.
Co-founder Jack Dorsey described the experience as feeling like a conversation with a person, while project lead Ragavan Thurairatnam called it a “GPT3 moment for speech.” Those are statements from people involved with the project, not independent findings about the model’s abilities. The available assessment remains limited to demonstrations selected and published by Block and the reactions of journalists who listened to them.
Other details remain unanswered. TechCrunch reported that it was unclear whether emotional qualities were influenced by tags, and CoinDesk noted the absence of model size, benchmark results and a release date. Block has also framed learning new languages as an open question. The Decoder cautioned that sounding emotional is not evidence that a system feels emotion or has human-like intelligence.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Jack Dorsey, Block co-founder
Testing and Release Details Remain Open
Dessa has not been made available for public testing, and the source material reports no independent access beyond a small early-access group of AI safety researchers. Without a technical paper, model card, benchmarks or public weights or API, outside reviewers cannot verify Block’s claims about training data, performance, speed or hardware requirements.
It is also unclear which languages the model supports, whether it can imitate a specific person, and what safeguards or consent checks it uses. Block has not stated whether the model includes watermarking. The examples show how selected conversations sound, but they do not establish how the system performs across ordinary use, difficult conditions or attempts to misuse it.
Nor do the demonstrations establish that the system experiences emotions. A model may produce speech that sounds amused, concerned or grateful without feeling those states. That distinction remains relevant to both consumer expectations and any claims made about the technology.
Watch for Independent Access
The next meaningful evidence would be independent access and published technical details: a paper or model card, documented evaluations, clearer release plans and information about consent and safeguards. Block has not announced a public release date, so it is not yet clear when outside researchers or customers will be able to test Dessa themselves.
Until those details emerge, readers should treat the demos as examples of what Block says its research model can do—not proof that all the capabilities work reliably or that the system is ready for deployment. Any future release could clarify performance and safety questions, but the current announcement leaves them unresolved.
Key Questions
What is Block’s Dessa?
Dessa is a speech-to-speech research model that Block says can listen and speak at the same time. The company says it was trained from scratch on recorded human conversation.
Has anyone outside Block tested the model?
The available source material reports that no outsider has directly tested Dessa. Journalists have listened to Block’s selected demo recordings, and a small group of AI safety researchers has been offered early access.
Can Dessa be used to clone a specific person’s voice?
That has not been established. Block has not clarified whether the model can imitate a specific person or described its consent checks and safeguards in the source material.
Does Dessa show that AI can feel emotion?
No. The demos may sound expressive, but an emotional-sounding voice is not evidence that a model experiences emotion. The distinction was also raised by The Decoder in its coverage.
When will Dessa be publicly available?
Block has not announced a public release date, and it has not released model weights or an API, according to the source material available as of October 26, 2026.
Source: ThorstenMeyerAI.com
