🔍 Read the full analysis: Can You Tell What’s Human By Listening? Block’s Dessa Raises The Question 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 trained on recorded human conversation, and published demonstrations of it responding with laughter, pauses and interruptions. The model is not publicly available, and independent testing, technical documentation and details about safeguards remain unavailable.
Block published demonstrations of Dessa, a speech-to-speech research model that the company says was trained from scratch on recorded human conversation, on October 23. The clips show it laughing, interrupting and responding to emotional cues, but the model is not publicly available and independent researchers have not been able to test the claims—leaving open questions about its capabilities and potential use in voice fraud.
Block says Dessa was trained using human conversation alone, without text or pretrained models, and can listen while speaking. The company describes it as able to recognize and produce signals such as sarcasm, laughter and pauses, and says it runs faster than real time on a MacBook Pro M4 Max or better. Block has not given a specific latency figure or published a technical paper, model card or benchmark results.
The recordings on Block’s project page include a laugh mid-sentence, a conversation in which Dessa helps recall a film title, and a response to an anniversary surprise. Block says the clips are live conversations and were not steered. That description is the company’s account; outsiders have not been given access to the model to reproduce the demonstrations or assess how representative they are.
External coverage has been more qualified than the launch claims. The Verge described some exchanges as “a little awkward and uncanny,” while saying Dessa came across as more natural than another model featured in the comparisons. TechCrunch reported that it was unclear whether emotional expression was guided by tags. Block has offered access to a small group of AI safety researchers, but has announced no general release, public API or model weights.
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 Natural Voices Meet Fraud
Dessa’s demonstrations matter because conversational realism is relevant both to helpful voice interfaces and to impersonation scams. A system that responds to interruptions and emotional cues could make automated calls feel less mechanical. That does not establish that Dessa can clone a particular person or that it has been used in fraud; Block has not disclosed whether it can imitate specific voices.
The risk exists independently of this model. The FBI’s 2025 Internet Crime Report, published in April 2026, counted 22,364 complaints mentioning AI, with $893 million in reported losses. The FBI also said distress scams using cloned voices of loved ones accounted for more than $5 million in losses. These figures describe reported complaints and losses, not proof that Dessa contributed to them or a measure of its future impact.
For businesses and families, the practical point is that a familiar-sounding voice alone is not reliable proof of identity. Payment requests should be verified through a separate, known contact method. Families can agree on a code word and call back using a number they already trust. Those are general precautions, not responses to a demonstrated misuse of Dessa.
Products Worth Considering
Canon PIXMA PRO-200S Wireless Professional Inkjet Photo Printer
Logitech MX Mechanical Full-Size Wireless Mechanical Keyboard, Tactile Quiet Switches, Backlit Keys, Bluetooth, USB-C, Multi-Device, Graphit
Canon imagePROGRAF PRO-1100 Professional 17-Inch Wireless Inkjet Photo Printer Bundle
As an affiliate, we earn on qualifying purchases.
Block’s Speech-Only Research
Block, the company behind Square and Cash App, introduced the project on October 23. Co-founder Jack Dorsey wrote that talking with Dessa “feels like talking with a person,” describing its emotion, nuance, pauses and cadence. That is his assessment of the demonstrations, not an independent evaluation.
Block says the research team trained the model using one NVIDIA GB200 NVL72 and a comparatively small amount of language data. The company has not supplied the model’s size, training-data details or benchmarks, so those claims cannot yet be compared independently with other speech systems. The project’s name refers to a Toronto deep-learning startup acquired by Square in 2020, according to CoinDesk.
Block frames the work as an effort to build voice AI that can respond to the emotional qualities of speech, which it says conventional speech-to-text systems can lose. Dessa’s lead, Ragavan Thurairatnam, called it “a GPT-3 moment for speech” and said it could run on consumer GPUs. The model’s actual hardware requirements and performance outside Block’s reported setup have not been independently established.
““Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.””
— Jack Dorsey, Block co-founder
What Testing Has Not Established
As of October 26, no public model, weights, API, technical paper or benchmark results were available. Reporters and outside researchers have not been able to independently test the system. It is therefore unclear how often the behaviors in Block’s selected clips occur, how it performs in longer or less controlled conversations, and how it compares with other speech models under consistent conditions.
Block has not publicly detailed the training data, languages supported, safeguards, watermarking or consent checks. The company says Dessa chooses its own voice and personality, but has not clarified whether the model can imitate a named individual or how it would prevent that. The demonstrations establish that Block has produced the clips it published; they do not establish general reliability, safety or vulnerability to a particular fraud technique.
Access and Safeguards to Watch
Block says a small group of AI safety researchers has early access, while other interested users can join a waitlist. The company has not provided a public release date. Wider access or independent evaluations could clarify whether the demonstrations hold up outside Block’s own recordings and how the model handles misuse risks.
Until further documentation or testing is available, the key developments to watch are publication of a paper or model card, details about supported languages and data practices, and evidence about safeguards against impersonation. For now, claims about Dessa’s performance should be treated as company claims, while voice-based identity checks remain vulnerable in principle to existing cloning technology.
Key Questions
What is Block’s Dessa?
Dessa is a speech-to-speech research model that Block says was trained from scratch on recorded human conversation. The company says it can listen and speak at the same time.
Can the public try Dessa?
Not currently. Block has announced a waitlist and limited early access for a small group of AI safety researchers, but no general release date, public API or model weights.
Has anyone independently verified Block’s claims?
Not through direct access to the model, based on information available on October 26, 2026. Reviews have assessed Block’s published clips, but no public benchmark or independent model test has been reported.
Does Dessa clone a specific person’s voice?
Block has not said whether Dessa can imitate a specific person, or described related consent and anti-impersonation safeguards. Its ability to do so is not established by the published demonstrations.
How can people check a suspicious voice call?
Do not rely on a familiar voice as proof of identity. For an urgent request for money, hang up and call the person back using a number you already know; families may also agree on a private code word.
Source: ThorstenMeyerAI.com
