🔍 Read the full analysis: Block’s Dessa Brings Human-Like Conversation To AI. What Can We Infer? on ThorstenMeyerAI.com
Get hardware and tech essentials 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 whose demos show it laughing, responding to emotion and handling interruptions. The demonstrations are company-selected; no paper, public model, independent testing or release date has been announced. Its human-like voice may improve conversation, but it also sharpens concerns about impersonation and voice-based fraud.
Block has published demonstrations of Dessa, a speech-to-speech research model that the company says can listen while speaking, respond to emotion and interrupt naturally. The recordings suggest a more human-like conversational style, but the model has not been made publicly available or independently tested, leaving its actual performance and safeguards unknown.
Block’s project page says Dessa was trained from scratch on recorded human conversation alone, without text or pretrained models. The company says the model processes speech directly, rather than converting it to text first, and can respond to cues such as laughter, sarcasm or a frustrated sigh. It also says Dessa can listen while talking and decide when to speak or provide conversational feedback.
The demonstrations include a laugh mid-sentence, a response to someone struggling to remember a film title, and a conversation about an anniversary surprise. Block describes the clips as live recordings and says nobody steered the model. They remain Block-selected examples, not a public test set. The Verge reported that the conversations could sometimes feel awkward or uncanny, while TechCrunch said it was unclear whether emotional delivery might be guided by tags.
Block says inference runs faster than real time on a MacBook Pro M4 Max or better, but has not provided a latency figure. The company also says training used one NVIDIA GB200 NVL72. No technical paper, model card, benchmark results, license, public weights or API have been announced. Block says access is limited to a waitlist and a small group of AI safety researchers.
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 a Voice Sounds More Convincing
Dessa matters because conversational realism has both practical and safety implications. A system that responds to interruptions and emotional cues could make voice assistants easier to talk with. The same qualities could make a fraudulent call harder to recognize by ear. That is a risk assessment, not evidence that Dessa has been used in fraud.
The source material 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 reported more than $5 million in losses from distress scams using cloned voices of loved ones. Those figures cover reported cases and do not establish that conversational models such as Dessa caused them.
The immediate takeaway is that a familiar-sounding voice should not be treated as proof of identity, particularly when a caller demands urgent payment or sensitive information. Organizations and families can use independent callback numbers and a prearranged verification phrase. For companies deploying synthetic voices, disclosure and consent questions also matter; the source says Article 50 of the EU AI Act applies from August 2, 2026. The details of how that rule applies to specific uses depend on the circumstances.
Products Worth Considering
Apple MacBook Pro 14 (2021) with M1 Max, 64 GB RAM, 2 TB SSD, Silver
Apple MacBook Pro 14.2-inch with M5 Max, 18-Core CPU, 20-Core GPU, 48GB Memory, 2TB SSD, Space Black
Apple MacBook Pro 2021 16‑inch with M1 Max, 32 GB RAM, 2 TB SSD, Space Gray (QWERTY English)
As an affiliate, we earn on qualifying purchases.
What Block Has Released So Far
Block, the company behind Square and Cash App, published the Dessa project page on October 23, 2026. Co-founder Jack Dorsey described talking with the system as like talking with a person, citing emotion, pauses and natural cadence. Dessa lead Ragavan Thurairatnam called the work “a GPT3 moment for speech” and said it could run on consumer GPUs. These are Block representatives’ assessments, not independently established performance findings.
The model’s name refers to a small research lab at Block founded by former members of Toronto-based deep-learning startup Dessa, which Square acquired in 2020, according to the source material. Block’s demos also compare Dessa with another company’s system, GPT-Live-1. Reporters found the model could sound more natural in some exchanges, but the clips do not establish how it performs across users, subjects, accents or difficult conversational conditions.
Block presents direct speech processing as a way to preserve vocal information that speech-to-text systems may discard. That could help a system respond to how a person sounds, not just the words they say. But expressive speech is not evidence of human feeling, and a handful of recordings cannot establish reliability, broad usefulness or safety.
“Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.”
— Jack Dorsey, Block co-founder
The Model Has Not Faced Outside Tests
As of October 26, Block had not released a technical paper, model card, benchmark results, license, public weights or API. That means outside researchers cannot verify the company’s claims about training, conversational ability, speed or hardware requirements. The demos show what happened in selected recordings; they do not establish consistent performance.
Several safety and access questions also remain unanswered. Block has not disclosed which languages Dessa supports, whether its voice can imitate a particular person, or what watermarking, consent checks or other safeguards it uses. The company says Dessa chooses its own voice and personality, but that statement does not resolve how the model might behave when adapted or deployed. A public release date is also unknown.
It is not clear whether the model’s emotional qualities emerge from training alone or are partly guided by controls such as tags. Nor is there evidence in the source material that Dessa has been used to deceive callers or that it can reliably pass as a specific person. Potential misuse should not be reported as demonstrated misuse.
Release Details and Safety Tests
The next meaningful developments would be a technical report, independent evaluations and clearer information about access. Those could let researchers examine whether the model works as Block describes across different speakers and situations, and test how it handles interruptions, emotional cues and attempts to imitate a person. Block has announced no public release date, so when those tests may be possible is unknown.
Until more information is available, users should verify unexpected requests through a number they already trust rather than calling back the number on an urgent incoming call. Families can agree on a private code phrase for emergencies. Businesses can use a second verification channel for payment instructions and avoid relying on voice alone for identity checks. These are general precautions against impersonation; they do not depend on Dessa becoming public.
For Block, outstanding questions include what safety testing will be shared, whether synthetic speech will be marked or disclosed, and how the company will govern access beyond its current limited group. The project remains a research demonstration, and its capabilities and risks cannot yet be assessed from public evidence alone.
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 and respond to vocal cues such as laughter and emotion. It has not been publicly released.
Has anyone independently tested the model?
Not publicly, based on information available as of October 26, 2026. Reporters have assessed Block’s recordings, but there are no published benchmarks or independent tests of the model itself.
Does the announcement show that Dessa is being used for scams?
No. The source material provides no evidence that Dessa has been used in fraud. The concern is that more natural speech could make voice impersonation harder to detect, while voice-cloning scams are already reported.
When will the public be able to use Dessa?
Block has not announced a public release date. The company says access is currently limited to a waitlist and a small group of AI safety researchers.
Source: ThorstenMeyerAI.com
