🔍 Read the full analysis: The New AI Voice: Block’s Dessa Laughs, Hesitates, And Interrupts 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 on October 23, describing it as a speech-to-speech model trained only on recorded human conversation. Company-selected demos show laughter, pauses and interruptions, but outsiders have not tested the model and Block has released no paper, benchmarks, weights or API. Its humanlike voice also raises questions about impersonation and safeguards.
Block announced Dessa, a speech-to-speech research model that the company says can laugh, respond to emotion and interrupt naturally, releasing selected audio demonstrations on October 23. The demos offer a glimpse of a more conversational voice interface, but the model has not been made publicly available, and its performance and safeguards have not been independently verified.
Block says Dessa was trained from scratch on recorded human conversation alone, without text or pretrained models. The company describes it as a full-duplex system: it can listen while speaking and respond with conversational cues such as backchanneling or cutting in. Its demonstrations include laughter during a sentence, a listener helping retrieve a forgotten film title, and a response to an anniversary surprise.
According to Block, inference runs faster than real time on a MacBook Pro M4 Max or better. The company also says training used one NVIDIA GB200 NVL72 and hundreds of times fewer words than large language models require. It did not provide a latency figure, disclose the model’s size or publish benchmark results. Those specifications remain company claims, not independently tested findings.
Block says each published clip is a live recording and that nobody steered the model during the conversations. The Verge found the interactions more natural than a comparison model in some respects, while describing moments as awkward or uncanny. TechCrunch reported that it was unclear whether emotional delivery was directed by tags. No outside researcher has run Dessa, and access is limited to a waitlist and a small group of AI safety researchers, according to the source material.
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
Humanlike Voices Raise Fraud Concerns
A voice that responds to interruptions and conveys emotion could make automated calls feel more natural. That may improve conversational software, but it may also make it harder for a listener to distinguish a generated voice from a person. The demonstrations do not establish that Dessa can clone a specific individual; Block has not said whether the model can imitate a named person.
The risk is relevant because voice impersonation scams already exist. The FBI’s 2025 Internet Crime Report, released in April 2026, counted 22,364 complaints mentioning AI and $893 million in reported losses across those complaints. It also recorded more than $5 million in losses from distress scams using cloned voices to imitate relatives. These figures cover reported cases and do not show that Dessa was involved.
For organizations, the development is a reminder that a familiar voice alone is weak proof of identity, particularly when a caller requests a payment or sensitive information. A callback to a known number and a second verification method can reduce reliance on voice recognition. Families can agree on a private code word and verify urgent requests through a separate channel. Urgency should prompt verification, not immediate action.
Products Worth Considering
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Apple Mac mini Desktop Computer with M5 Pro Chip (2026)
Brother TD-2020A 2‑Inch Direct Thermal Desktop Label Printer with USB
As an affiliate, we earn on qualifying purchases.
What Block Has Released So Far
Block, the company behind Square and Cash App, published a project page for Dessa on October 23. Co-founder Jack Dorsey described the conversations as having emotion, nuance, pauses and natural cadence. Dessa is associated with a small Block research lab founded by former members of Toronto-based AI company Dessa, which Square acquired in 2020, according to CoinDesk.
The launch material describes a research project, not a product release. As of October 26, the source material reports no technical paper, model card, public weights, API or release date. Block has not listed supported languages or explained whether the system uses watermarking or consent checks. The company has framed learning new languages as an open question.
Outside coverage has added caution to the launch claims. The Decoder noted that emotional-sounding speech is not proof of emotion or intelligence. Block’s argument for the research is that voice systems can miss emotional cues, such as pain or distress. The same ability to detect and express cues, however, can make a synthetic interaction feel more convincing. The demos show a capability claim, not a complete safety assessment.
“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 Safeguards Remain Unknown
There is not yet enough public evidence to judge how reliably Dessa handles ordinary conversations, noisy calls, accents, interruptions or longer exchanges. Block has supplied no public benchmark, technical documentation or model access. Independent performance testing remains unavailable.
Important safety questions are also unanswered in the available material: whether the system can reproduce a particular person’s voice, how it handles consent, and whether generated speech will be marked or watermarked. The number of languages it supports is not listed. Block’s statement that Dessa chooses its own voice and personality does not by itself resolve those issues.
It is also unclear when broader access might begin or whether Block plans a commercial release. The company’s selected recordings cannot establish how the model behaves across users and settings. Claims about realism, speed and training efficiency should be treated as unverified until more evidence is published.
Awaiting Public Research and Access
The next concrete developments would be a technical paper, model documentation, independent testing or a public release date. Those materials could clarify the system’s capabilities, training approach, supported languages and safety measures. Block has not announced a timetable for any of them.
Until that information appears, Dessa is best understood as a research model demonstrated through company-selected recordings, not a tool available for general use. Companies and households do not need to wait for its release to adopt basic verification practices: confirm unusual payment requests through a known channel, and do not treat a familiar-sounding voice as proof of identity. Whether Block will add safeguards or open the model to external review remains to be seen.
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’s demos show conversational cues such as laughter and interruptions.
Can the public use Dessa now?
No public release, weights or API have been announced in the source material. Block says access is through a waitlist and a small group of AI safety researchers.
Has Dessa been independently tested?
No outside researcher has run the model, according to the source material. Current evidence consists of Block-selected recordings and reporting by outlets that listened to those demonstrations.
Does Dessa clone a specific person’s voice?
That is not established. Block has not clarified whether Dessa can imitate a specific person or described consent checks or watermarking in the available material.
How can people verify a suspicious call?
Hang up and call the person back using a number already known to you. Families can also agree on a private code word; organizations should use a separate verification method for payment requests.
Source: ThorstenMeyerAI.com
