🔍 Read the full analysis: AI Voices That Laugh And Hesitate: What Block’s Dessa Means For Listeners on ThorstenMeyerAI.com
Get business pricing on office and shipping supplies
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
TL;DR
Block announced Dessa, a speech-to-speech research model that it says can listen and respond in real time, with laughter, pauses and other emotional cues. The demonstrations are company-provided and the model has not been independently tested or publicly released, leaving its capabilities and safeguards unverified.
Block published demonstrations on October 23 of Dessa, a research model it says can listen and speak at the same time while conveying cues such as laughter, pauses and frustration. The recordings suggest a more conversational style of voice AI, but the model has not been independently tested or released publicly, so its performance and safeguards remain unverified.
Block says Dessa was trained from scratch on recorded human conversation, without text or pretrained models. The company says the system can interrupt or respond with backchannel cues while a person is speaking, rather than waiting for a spoken exchange to finish before generating a reply. Block also says it runs faster than real time on a MacBook Pro M4 Max or better, but has not provided a latency measurement.
The project page includes recordings of Dessa laughing mid-sentence, responding to a surprise and helping a speaker recall a film title. These are Block-selected demonstrations, not results from independent tests. The Verge described some exchanges as “a little awkward and uncanny,” while reporting that Dessa sounded more natural than another model used in a comparison. TechCrunch noted that it was unclear whether emotional delivery might be steered by tags.
Block co-founder Jack Dorsey wrote that talking with the model feels like talking with a person, citing its “emotion, nuance, pauses, and a natural cadence.” Dessa lead Ragavan Thurairatnam called it “a GPT3 moment for speech,” according to the source report. Those are assessments by people involved in the project, not independently established measures of capability.
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
Natural Speech Raises Impersonation Risks
More fluid speech could make voice assistants easier to use, especially in conversations where people interrupt, hesitate or rely on tone to communicate. Block argues that models able to recognize emotion could respond more appropriately when a person sounds distressed. That potential benefit has not yet been demonstrated outside the company’s recordings.
The same conversational realism could make fraudulent calls harder to recognize by ear. The source 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 lost more than $5 million to distress scams using cloned voices. These figures describe reported complaints and losses for that period; they do not show that Dessa has been used in fraud or that it changes the scale of such scams.
For listeners, the practical point is that a familiar-sounding voice should not be treated as proof of identity. A caller’s apparent warmth, hesitation or emotion can be generated or imitated. In higher-risk situations—such as requests for money or account access—verification through a separate, known channel matters more than how convincing the call sounds.
Products Worth Considering
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Canon PIXMA PRO-200S Wireless Professional Inkjet Photo Printer
Canon imagePROGRAF PRO-1100 Professional 17-Inch Wireless Inkjet Photo Printer Bundle
As an affiliate, we earn on qualifying purchases.
A Research Preview, Not a Public Product
Block, the company behind Square and Cash App, posted the Dessa project page on October 23, 2026. The project is being developed by a small research lab that Block says was founded by former members of the Toronto deep-learning startup Dessa, which Square acquired in 2020, according to CoinDesk. The name refers to the research project, not a generally available voice service.
As of October 26, access was described as a waitlist, with early access for a small group of AI safety researchers. Block had not announced model weights, an API or a public release date. News organizations and other outside users therefore could not test the system directly. The public evidence consisted of demonstrations selected and published by Block.
The source report also notes that Block has not published a technical paper, model card, benchmark results or license. The company says the system was trained using one NVIDIA GB200 NVL72 and learned from fewer words than large language models typically require, but detailed methods and measurements were not provided. Those statements remain company claims until technical documentation or independent evaluations are available.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Block co-founder Jack Dorsey
Capabilities and Safeguards Remain Untested
It is not yet clear how Dessa performs across longer conversations, noisy environments, different speakers or languages. Block has not released benchmarks, a model card or a technical paper, and outside researchers have not had general access to test its claims. The demonstrations cannot establish how reliably it handles interruptions or whether its emotional cues are consistent.
Other open questions include whether the model can imitate a specific person, how it selects its voice and personality, and whether outputs will be watermarked or subject to consent checks. Block says the model picks its own voice and personality, but the source material does not identify independent safeguards or provide evidence about misuse resistance. The available information also does not establish whether the model’s apparent emotional responses reflect anything beyond generated speech.
Awaiting Independent Access and Evidence
The next concrete milestone is whether Block expands access beyond its waitlist and small safety-research group. A public release, technical documentation or independent evaluations would help establish the model’s actual performance, operating limits and measures against impersonation. Block had not announced a release date as of October 26.
Until more evidence is available, listeners and organizations can treat Dessa as a research preview rather than a deployed product. For calls involving urgent requests for money or sensitive information, people can hang up and call back using a number they already trust, or agree on a family code word for emergencies. Organizations can avoid relying on voice alone to authorize payments and can require a second verification method. These are general precautions against voice impersonation; the source does not report that Dessa itself has been used in scams.
Key Questions
What is Block’s Dessa?
Dessa is a speech-to-speech research model that Block says was trained on recorded human conversation. The company says it can listen and speak at the same time and produce conversational cues such as laughter and pauses.
Can the public use or test Dessa now?
Not generally, based on the status reported on October 26, 2026. Block offered a waitlist and limited early access for a small group of AI safety researchers, but had not announced public model weights or an API.
Has Dessa been shown to make voice scams more common?
No. The available material does not show that Dessa has been used in fraud or measure any effect it might have on scam rates. The broader concern is that more natural-sounding synthetic speech could make voice impersonation more convincing.
What should I do if a familiar voice calls asking for money?
Do not rely on the voice alone to confirm identity. Hang up and call back using a number you already know, and consider agreeing on a family code word for emergencies.
Source: ThorstenMeyerAI.com
