🔍 Read the full analysis: Why Block’s Dessa Sounds Human—and Why Voice Alone Proves Nothing on ThorstenMeyerAI.com
Get movie nights delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
TL;DR
Block announced Dessa, a speech-to-speech research model that the company says can listen and respond in real time, including with laughter and interruptions. Its human-like voice has only been shown in Block-selected recordings: there is no public model, technical paper or independent testing, and realistic voices can complicate defenses against impersonation scams.
Block has published demonstrations of Dessa, a speech-to-speech research model that the company says can respond while listening, laugh and handle conversational interruptions. The recordings suggest a more natural-sounding voice interface, but no outside researcher or reporter has tested the model directly, and Block has not released a paper, model card or public access. That gap matters because realistic voices can also be used to impersonate people on calls.
Block’s project page, published October 23, describes Dessa as trained from scratch on recorded human conversation, without text or pretrained models. The company says it can hear and produce emotional cues such as sarcasm, laughter and a sigh, while listening and speaking at the same time. Block also says it runs faster than real time on a MacBook Pro M4 Max or better; it has not provided a specific latency measurement.
The examples on Block’s page include a laugh during a sentence, a conversation about a forgotten film title and a response to an anniversary surprise. Block says the clips are live recordings and that nobody steered the conversations. Those are company-provided demonstrations, not independent tests. The Verge described some exchanges as “a little awkward and uncanny,” while saying Dessa came across as more natural than another model used for comparison. TechCrunch reported that it was unclear whether emotional delivery was being guided by tags.
Block says the model was trained using one NVIDIA GB200 NVL72 and learned from far fewer words than large language models typically use. Those statements, like its claims about performance, have not been independently verified in public testing. The company has disclosed no model size, benchmark results, weights, API or release date. Access is limited to a waitlist and early access for a small group of AI safety researchers, according to the source report.
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
Human-Sounding Calls Raise Fraud Risks
Dessa does not need to be publicly available to highlight a broader security issue: a convincing voice is not proof of identity. Voice cloning has already appeared in distress scams, where a caller imitates a relative, and in attempts to persuade businesses to transfer money. A system that handles pauses and interruptions naturally could make a conversation feel more ordinary, but Block’s demos do not establish how well Dessa imitates a particular person or whether it can be used for fraud.
The FBI’s 2025 Internet Crime Report, released in April 2026, counted 22,364 complaints mentioning AI and $893 million in reported losses. The report also recorded more than $5 million in losses from distress scams using cloned voices. These figures concern AI-related complaints and reported losses across the report’s categories; they do not measure losses caused by Dessa. The figures do show why organizations and families should not treat voice familiarity as a reliable check when money or urgent requests are involved.
For companies, the practical implication is to verify sensitive requests through a second channel, such as calling back on a known number. For families, a shared code word and a pause before sending money can help counter pressure tactics. These precautions address voice impersonation generally; the source material does not establish that Dessa itself has been used in a scam.
Popular right now
Bose QuietComfort Headphones (2nd Gen), Wireless Noise-Cancelling Over-Ear Headphones, Eucalyptus Green
TCL 85 Designer Series QD-Mini LED TV (Model 85A400U) with Google Smart TV
Philips 4400 Series Fully Automatic Espresso Machine with LatteGo Milk System
As an affiliate, we earn on qualifying purchases.
What Block Has Demonstrated
Dessa comes from a small research lab at Block, the company behind Square and Cash App. The source report says the lab was founded by former members of Toronto-based AI startup Dessa, acquired by Square in 2020. Block’s co-founder Jack Dorsey wrote that speaking with the model feels like talking with a person, citing its emotion, pauses and cadence. That is Dorsey’s assessment, not an independent evaluation.
The model’s design differs from systems that convert speech to text, process the text and then generate spoken output. Block says Dessa works directly with audio and can respond while the other person is still talking. The company argues that this lets it respond to emotional signals that a text-based pipeline may miss. The same conversational fluency, however, can make it harder for a listener to tell whether a voice is synthetic.
Outside coverage has stressed the limits of the available evidence. CoinDesk reported that Block had not disclosed the model’s size, benchmark results or a public release date. The Decoder noted that sounding emotional does not establish that a system feels emotion or has human-like understanding. As of October 26, the source report found no technical paper, model card, license or stated list of supported languages.
“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 Are Still Missing
The central question is how Dessa performs outside curated examples. No outsider has run the model, and the public material does not provide benchmarks, a technical paper or enough detail to reproduce its results. The demonstrations cannot establish whether it reliably handles interruptions, responds appropriately across different accents, or maintains the same quality in longer conversations.
It is also unclear which languages it supports, whether it can imitate a specific person, and what safeguards it has for consent or watermarking. Block says Dessa chooses its own voice and personality, but the source material does not describe technical controls that would prevent misuse. Nor does it establish that the model has been used in voice fraud. Those distinctions matter: potential risk is not evidence of an actual incident, and a convincing demo is not proof of broad capability.
Release Details and Safeguards Awaited
Block has not announced a public release date, model weights or an API. The next concrete steps for evaluating Dessa would be independent access, technical documentation and testing that reports performance across varied speakers and situations. Details about supported languages, consent protections and whether outputs can be identified as synthetic would also help clarify the model’s capabilities and risks.
Until that information is available, the demonstrations should be treated as evidence of what Block chose to show, not proof of how the system performs in general. For people and organizations facing urgent requests by phone, the immediate response does not depend on Dessa: hang up and call back using a trusted number, and verify requests for money through a separate channel.
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. Block announced it on October 23, 2026.
Can the public use or test Dessa?
Not broadly, based on information available as of October 26. Block has described a waitlist and limited early access for AI safety researchers, but has not announced public weights or an API.
Has Dessa been proven to imitate a specific person?
No public evidence in the source material establishes that. Block says the model chooses its own voice and personality, but has not publicly detailed whether it can imitate a person or what safeguards prevent that.
Does a human-sounding voice prove that a caller is real?
No. A familiar or natural-sounding voice is not reliable identity verification. For urgent requests involving money, contact the person or organization through a number or channel you already trust.
Source: ThorstenMeyerAI.com
