🔍 Read the full analysis: Block’s Dessa Laughs, Hesitates And Interrupts Like A Person. Why A Human-Sounding Voice Is No Longer Proof Of Anything on ThorstenMeyerAI.com
Get business pricing on tech for your team
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Block introduced Dessa, a speech-to-speech research model whose company demos feature laughter, interruptions and emotional vocal cues. The model is not publicly available, and no independent evaluation, technical paper or benchmark has been published, so its capabilities and safeguards remain unverified.
Block published demonstrations of Dessa, a speech-to-speech research model that the company says can laugh, respond to emotion and interrupt naturally, but the system has not been made available for independent testing. The release matters because more realistic voice interactions could make customer service and other voice applications feel less mechanical, while also making it harder for people to distinguish a synthetic caller from a real person.
Block says Dessa was trained from scratch on recorded human conversation, without text or pretrained models. The company describes it as a full-duplex system: it can listen while speaking, and is designed to respond to pauses, interruptions and conversational cues such as laughter or sarcasm. Those are claims from Block’s project materials, not findings from an outside evaluation.
The company says inference runs faster than real time on a MacBook Pro M4 Max or better, but has not supplied a latency figure. Block also says the model was trained using one NVIDIA GB200 NVL72. Its published examples include a laugh mid-sentence, a conversation about a forgotten film title and a response to an anniversary surprise. The recordings were selected and released by Block.
As of October 26, Block had published no technical paper, model card, benchmark results, weights or API, according to the source report. Access was described as limited to a waitlist and a small group of AI safety researchers. The Verge found some exchanges “a little awkward and uncanny,” while TechCrunch said it was unclear whether emotional responses might be guided by tags. Neither outlet had independently run the model.
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
Dessa’s conversational style could matter beyond the quality of a demo. Voice cloning is already used in scams, and a system that handles interruptions, laughter and emotional cues could make synthetic conversations more convincing. That does not establish that Dessa can clone a particular person or has been used for fraud; Block has not publicly confirmed either point.
The source report cites the FBI’s 2025 Internet Crime Report, published in April 2026, which recorded 22,364 complaints mentioning AI and $893 million in reported losses. It also says victims reported more than $5 million in losses from distress scams involving cloned voices of loved ones. These figures describe reported complaints and losses across AI-related cases; they do not measure Dessa’s role or the effect of this particular model.
For organizations, the development is a reminder not to treat a familiar-sounding voice as proof of identity. Payment requests and sensitive account changes can be checked through a separate, known contact channel. For families, a pre-agreed code word and calling back on a trusted number can help verify an urgent request. These precautions address voice impersonation generally; the source does not report a new scam involving Dessa.
Products Worth Considering
Jackery Explorer 300 Portable Power Station, 292Wh
Jackery Explorer 2000 v2 Portable Power Station with Two 200W Solar Panels
Bose SoundLink Flex Bluetooth Speaker (2nd Gen), Portable Waterproof Speaker, Black
As an affiliate, we earn on qualifying purchases.
Block’s Speech-Only Research Claims
Block, the company behind Square and Cash App, announced Dessa on October 23. The project is associated with a small research lab whose members previously worked at Dessa, a Toronto deep-learning startup acquired by Square in 2020, according to CoinDesk. Block presents the system as an exploration of speech that retains vocal information—such as emotion and timing—that can be lost when speech is converted to text and then processed by a language model.
Co-founder Jack Dorsey described talking with the model as feeling like talking with a person, citing emotion, nuance, pauses and natural cadence. Lead Ragavan Thurairatnam called the project “a GPT3 moment for speech.” Those are attributed opinions from people involved in the launch, not independent evidence of performance. The company’s demonstrations are the only public examples described in the source material.
Reaction has included both interest in more natural voice interfaces and concern about deception. The source report notes that Block frames emotional awareness as useful for responding appropriately to people in pain, while critics question why AI voices should sound human. The central dispute is not whether the demos contain human-like cues, but how consistently the model produces them, how it is controlled and what protections accompany access.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Jack Dorsey, Block co-founder
Independent Tests and Safeguards Missing
Dessa’s performance has not been independently verified, because no public model, API or weights are available. Block has not published a paper, model card, benchmarks, model size or release date in the material cited. Company-selected recordings cannot establish how the system performs across ordinary conversations, difficult conditions or extended use.
Other unresolved questions include which languages the model supports, whether its emotional style is directed by tags, and what protections it has against impersonation. The source report says Block has not provided public details about watermarking or consent checks, or confirmed whether Dessa can imitate a specific person. Block says the model chooses its own voice and personality, but that does not answer those questions. No evidence in the source establishes that Dessa has been used to deceive callers.
Release Details Will Shape the Test
The next meaningful milestone would be a technical paper, model card or wider release that lets independent researchers assess performance, limitations and safeguards. Block has not announced a public release date. The company’s stated access route remains a waitlist and a small early-access group of AI safety researchers, so the scope and timing of any external testing are unknown.
Until more information is available, Dessa should be treated as a research announcement backed by company demos, not a fully assessed or generally available product. Any future release will clarify whether its conversational realism holds up beyond selected clips and what measures Block puts in place around identity, consent and disclosure. The source report also notes that EU AI Act Article 50 applies from August 2, 2026, but does not establish how the rule applies to a particular use of Dessa.
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 and respond to conversational cues such as pauses, interruptions and emotion.
Can the public use or test Dessa?
Not broadly, based on the information available as of October 26, 2026. Block had announced no public weights or API and described access through a waitlist and a small group of AI safety researchers.
Have outside researchers verified Block’s claims?
No independent model testing is reported. Journalists listened to Block’s published clips, but the model itself was not publicly available for them to run.
Does Dessa clone specific people’s voices?
The available source material does not confirm that it can imitate a specific person. Block says the model chooses its own voice and personality, but its impersonation limits and consent protections remain unclear.
How can people check an urgent voice request?
For a request involving money or sensitive information, hang up and call back using a number already known to you. Families can also agree on a private code word. These are general precautions against voice impersonation, not responses to a reported Dessa-related scam.
Source: ThorstenMeyerAI.com
