🔍 Read the full analysis: Why Block’s Dessa’s Laughs And Pauses Don’t Prove It’s Human 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 that it says can listen and talk at once and convey emotion. Its demos are company-selected recordings; outsiders have not tested the model, and Block has not published a paper, benchmarks, weights or an API. Humanlike speech matters because it may make voice impersonation harder to spot, but the demos do not establish that Dessa can fool people or imitate a specific person.
Block published demonstrations of its speech research model Dessa on October 23, showing it laughing, pausing and responding during conversations. The clips illustrate the company’s claims about natural-sounding, simultaneous listening and speaking, but no outside researcher or reporter has tested the model directly, and Block has not released a paper, benchmark results or public access.
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, interrupt and provide conversational responses such as backchanneling. Its demonstrations include laughter, a sigh, a forgotten film title and a response to an anniversary surprise. These are recordings selected and published by Block, not independently reproduced tests.
Block also says the model runs faster than real time on a MacBook Pro M4 Max or better, and that training used one NVIDIA GB200 NVL72. The company has not supplied a latency figure or disclosed the model’s size, benchmark results, license or public release date. Dessa is available through a waitlist, with early access offered to a small group of AI safety researchers, according to the source material.
Outside coverage has offered a more qualified assessment than the launch claims. The Verge reported on October 23 that some conversations could feel awkward or uncanny, while finding Dessa more natural than another model in a comparison. TechCrunch noted October 24 that it was unclear whether emotion in the output might be guided by tags. Neither report involved public access to the model itself.
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
Why Natural-Sounding Calls Matter
A voice that laughs, pauses or reacts at the right moment may be more convincing than a flat automated voice. That could make it harder for someone on a phone call to rely on conversational oddities as a warning sign. But Dessa’s demos do not prove it can deceive listeners, and the source material does not establish that the model can copy a particular person’s voice.
The risk sits within a broader, already documented problem. 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 recorded more than $5 million in losses from distress scams involving cloned voices of loved ones. Those figures concern AI-related complaints and reported losses overall; they do not measure harm caused by Dessa.
For organizations, the development is a reminder that a familiar-sounding voice alone is not reliable proof of identity. Payment requests and sensitive changes should be checked through a separate, known channel. For families, a pre-agreed code word and calling back using a saved number can help verify a distress call. These precautions address existing voice-fraud risks, not a confirmed misuse of Block’s model.
Products Worth Considering
Sonos Era 100 SL Compact Microphone-Free Speaker, Black
LG UltraGear 27-Inch QHD OLED Gaming Monitor, 240Hz, 0.03ms, G-SYNC Compatible, FreeSync Premium Pro, HDMI 2.1
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
As an affiliate, we earn on qualifying purchases.
What Block Has Disclosed About Dessa
Block, the company behind Square and Cash App, posted its Dessa project page on October 23, 2026. Co-founder Jack Dorsey described the experience as “like talking with a person,” citing emotion, nuance, pauses and natural cadence. Dessa lead Ragavan Thurairatnam called it “a GPT3 moment for speech,” according to the source material. These are company and team characterizations, not independent performance findings.
The project page presents speech-only training and real-time interaction as central design choices. Block argues that common speech-to-text systems discard emotional information when they convert speech into text. It says Dessa can hear and produce features such as sarcasm, laughter and a voice cracking. Whether it does so consistently across different speakers, languages or difficult conditions has not been shown publicly.
Coverage from The Verge, TechCrunch and CoinDesk has noted limitations in the evidence, including the absence of independent testing and undisclosed technical details. As of October 26, the source material also reports no public technical paper, model card, weights or API. Block’s own project description identifies learning new languages as an open question.
What the Demonstrations Cannot Establish
Independent performance remains unverified. The available examples are Block-selected recordings, and the model has not been made available for general testing. The source material reports no technical paper, model card, benchmarks, public weights or API. Without those, outsiders cannot assess how consistently the model handles interruptions, whether its apparent emotion is produced through prompts or tags, or how it performs beyond the shown conversations.
Several safety and deployment questions are also unanswered. The source material identifies no Block statement on watermarking, consent checks or protections against imitating a specific person. Block says Dessa selects its own voice and personality, but that does not settle what users can do with the system or what safeguards will accompany any wider release. The company’s claims about on-device speed also lack a published latency measurement.
It is unclear when the waitlist will lead to broader access, which languages the model supports, or whether the early-access group can publish independent findings. No evidence in the source material shows that Dessa has been used in a scam or that it has caused financial losses.
The Tests Needed Before Wider Release
The next meaningful evidence would be independent testing and technical documentation covering benchmarks, latency, supported languages, model access and safeguards. Block has not announced a public release date in the source material, so the timing and terms of any broader rollout remain unknown. Results from the AI safety researchers with early access could add evidence if they are made public, but their access alone does not establish the model’s performance.
Until more information is available, readers should treat the demos as demonstrations of Block’s claims rather than proof that the model is indistinguishable from a person or capable of specific voice impersonation. Organizations can verify sensitive requests through established callback procedures, and individuals can pause urgent requests, hang up and call a known number. The FBI’s figures show that AI-related fraud complaints and losses were already reported in 2025, independent of what happens with Dessa.
Key Questions
What is Block’s Dessa?
Dessa is a speech-to-speech research model that Block says can listen and talk at the same time. The company says it was trained from scratch on human conversation, without text or pretrained models.
Has an independent group verified that Dessa sounds human?
Not in the material available as of October 26, 2026. Reporters have listened to Block’s published recordings, but no outsider has tested the model directly, and Block has not published benchmark results.
Does Dessa prove that voice calls can no longer be trusted?
No. The demonstrations do not establish that Dessa can fool listeners or copy a specific person. They do reinforce a broader caution: a familiar or natural-sounding voice should not be the only check for identity, especially for urgent money requests.
What should someone do if a family member calls asking for urgent money?
Pause the transaction, hang up and call the person back using a number already saved or independently verified. Families can also agree on a private code word for checking distress calls.
When will the public be able to use Dessa?
Block has not announced a public release date in the source material. Access is described as a waitlist, with early access for a small group of AI safety researchers.
Source: ThorstenMeyerAI.com
