🔍 Read the full analysis: Is A Human-Sounding Voice Still A Reliable Clue? Block’s Dessa 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 the company says was trained only on recorded human conversation and can respond with emotion, interruptions and backchanneling. The public evidence remains limited to Block-selected demos: the model has no public release, independent evaluation or published technical documentation, leaving its capabilities and safeguards unverified.
Block announced Dessa, a speech-to-speech research model, on October 23, showing recordings in which it laughs, responds to emotion and talks over or alongside a speaker. The company says the model was trained from scratch on human conversation alone, but no independent researcher has tested it and Block has not released a paper, benchmarks, model weights or an API. That gap matters as natural-sounding synthetic voices could make phone scams harder to identify by ear.
Block, the company behind Square and Cash App, describes Dessa as a system that listens and speaks at the same time, rather than converting speech to text before generating a reply. Its project page says the model can recognize and produce features such as sarcasm, laughter, pauses and a frustrated sigh. Block also says it was trained without text or pretrained models and can run faster than real time on a MacBook Pro M4 Max or better. The company gave no latency measurement to support that speed claim.
The demonstrations include Dessa laughing mid-sentence, helping a speaker recall a film title and responding to a story about an anniversary surprise. Block says each is a live recording and that “Nobody steers it.” Those recordings are still company-selected evidence; they do not establish how the system performs across users, languages, noisy calls or longer conversations. Block has described access as a waitlist and early access for a small group of AI safety researchers, not a general release.
External coverage has noted both the natural-sounding exchanges and their limits. The Verge described some conversations as “a little awkward and uncanny,” while reporting that Dessa sounded more natural than another model used for comparison. TechCrunch said it was unclear whether emotional effects were being guided by tags. CoinDesk reported that Block had not disclosed the model’s size, benchmark results or a public release date. No published technical evaluation is available to resolve those questions.
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 Stops Proving Identity
Dessa does not need to be publicly available for the announcement to matter: it illustrates the direction of speech generation, while existing voice-cloning tools already feature in reported fraud. A convincing voice can make an urgent call feel familiar, but familiarity is not authentication. Companies that approve payments or sensitive account changes based on a caller’s voice may need to rely on additional checks, such as a callback to a known number and a second approval channel.
The FBI’s 2025 Internet Crime Report, released in April 2026, counted 22,364 complaints mentioning AI and $893 million in reported losses associated with those complaints. It also reported more than $5 million in losses from distress scams using cloned voices of loved ones. These are figures for complaints and reported losses in the report’s stated categories; they do not measure Dessa’s impact or show that this model was used in any crime.
For families, the practical response is similarly basic: agree on a private code word, avoid sending money under pressure, hang up and call the person back using a number already saved or independently verified. Urgency itself is a warning sign. A natural voice may support a conversation, but it cannot establish who is speaking.
Popular right now
Apple AirPods Pro 3 Wireless Earbuds with Active Noise Cancellation
Sony WH-1000XM6 Wireless Noise-Canceling Over-Ear Headphones, Blue, Bundle
Breville Barista Pro Espresso Machine with Grinder, Brushed Stainless Steel
As an affiliate, we earn on qualifying purchases.
What Block Has and Has Not Published
Block’s project page appeared on October 23, 2026. Co-founder Jack Dorsey described speaking with the model as “like talking with a person,” pointing to emotion, pauses and natural cadence. Dessa lead Ragavan Thurairatnam called it “a GPT3 moment for speech” and said it could run on consumer GPUs. Those descriptions are the company’s characterizations, not results from a published independent study.
The research team is described as a small Block lab formed by former members of Toronto-based AI startup Dessa, which Square acquired in 2020, according to CoinDesk. Block says the model learned from hundreds of times fewer words than large language models require and was trained using one NVIDIA GB200 NVL72. It has not published enough technical detail for outside readers to verify the training comparison or reproduce the claimed performance.
Several details that would help assess the system remain absent: a technical paper, model card, license, public release schedule and list of supported languages. Block has described learning new languages as an open question. It also says Dessa chooses its own voice and personality, but that does not answer whether it can reproduce a specific person’s voice or what checks would prevent that. Those limits matter when evaluating both the demo and the model’s potential uses.
““Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.””
— Block co-founder Jack Dorsey
Independent Tests Still Missing
It is not yet clear how well Dessa performs outside the selected recordings, whether its emotional responses come from the model or additional steering, or how reliably it handles varied speakers and languages. Block has not disclosed benchmarks, model size or a public release date, and outsiders have not run the system. The company’s claims about its training setup and speed therefore remain unverified by public technical materials.
The available information also does not establish whether Dessa can imitate a named person, what consent or identity checks it uses, or whether generated audio is marked in a way listeners can detect. Block has not announced weights or an API. The demonstrations do not show that Dessa has been used in fraud, and no connection between this model and a specific scam is established by the source material.
A Release Could Answer Key Questions
The next useful milestones would be a technical paper or model card, independent evaluations, clearer information about supported languages and safeguards, and details on whether Block plans a broader release. The company has offered early access to a small group of AI safety researchers, but has not given a date for wider access. Until those materials appear, the demos show what Block chose to publish, not a complete assessment of Dessa’s capabilities or risks.
Businesses and families do not need to wait for this model to take precautions against voice impersonation. For financial instructions, use a separate verification step; for unexpected calls involving emergencies or money, end the call and contact the person through a trusted number. Do not treat a familiar-sounding voice as proof of identity.
Key Questions
What is Block’s Dessa?
Dessa is a speech-to-speech research model announced by Block on October 23, 2026. Block says it was trained from scratch using recorded human conversation and can listen and talk at the same time.
Has anyone independently tested the model?
No independent test results have been published in the source material. The evidence available so far consists of recordings selected and shared by Block, alongside journalists’ observations of those demonstrations.
Can Dessa clone a specific person’s voice?
That has not been established. Block says the system selects its own voice and personality, but has not publicly explained whether it can imitate an individual or what consent checks apply.
Does the announcement show that Dessa has been used in a scam?
No. The supplied reporting does not connect Dessa to a specific fraud. It discusses wider concerns about voice-cloning scams and cites FBI complaint and loss figures for 2025.
How can people check an urgent call asking for money?
Do not rely on voice recognition alone. Hang up and call back using a number you already trust, agree on a family code word, and verify payment requests through another channel before acting.
Source: ThorstenMeyerAI.com
