Safety
Dessa AI Voice Cloning: Can Dessa Copy Your Voice? What Block Has Said
It is one of the first questions people ask about any realistic voice AI: can it clone my voice? For Dessa, Block's speech-only model, the honest answer is narrower than the demos might suggest: Block has announced no voice cloning feature for Dessa — no tool that takes a recording of you and speaks in your voice, no named-person imitation, and no statement that Dessa can do it at all. This article separates what Dessa verifiably does with voices from what cloning would mean, and explains why Block's team, of all teams, treats the question carefully.
What Dessa does with voices, on the record
In Block's published demos, Dessa speaks in its own voice and personality — Block stresses that in its live, unsteered recordings the model picks what to say, when to say it and how, "down to its own voice and personality." What it demonstrably does is expressive: it laughs mid-sentence, gasps with what sounds like tearful gratitude, backchannels with "yeah" and "got it", adjusts its volume when asked, and hears emotion, sarcasm and frustration in the person talking to it. Those abilities come from its training — from scratch on human conversation audio only, no text, no pretrained models — as we explain in what Dessa AI is. Producing a convincing, emotional voice of its own is a different capability from replicating a specific person's voice, and Block has only ever demonstrated the first.
What voice cloning would mean — and why it is a different claim
Voice cloning means targeted imitation: feed a system samples of one person and have it say new things as that person. That is the capability behind the classic fraud scenario — a call that sounds like your boss, your bank or your grandchild. Nothing in Block's announcement claims Dessa does this. Block has published no cloning feature, no "upload a sample" tool, no list of voices it can imitate, and no demo of Dessa speaking as a named individual. The one demo that involves another voice at all is a conversation with OpenAI's GPT-Live-1 — two AIs talking to each other, each in its own voice, with Block noting OpenAI is unaffiliated. If you see a service today offering "Dessa voice cloning," it is not Block's: as our guide to Dessa scams and impersonation sets out, Block sells nothing, ships no app and has released no model files anyone could build a cloning tool from — there is no Dessa download and no open weights.
Why this team treats cloning carefully
The Dessa lab's own history explains the caution. The team descends from the Toronto startup Dessa, which became famous in 2019 for a synthetic Joe Rogan voice — a demonstration built to show how convincingly a specific person's speech could be imitated, and to warn about it early; our history of Dessa tells that story. Seven years later, the same lineage of researchers built a conversational model — and chose to give early access first to a small group of AI safety researchers, precisely to probe risks like impersonation before any wider release. Impersonation testing is part of the stated plan; whether a system like Dessa could be pushed toward cloning a target voice is exactly the kind of question those researchers are there to answer, and no findings have been published yet.
The practical takeaways today
First, you cannot clone a voice with Dessa, and nobody else can do it with Dessa either — the model is not public, not downloadable and not available through any API. Second, the general risk cloning represents does not depend on Dessa: other voice tools exist, so the durable defences are the same ones we recommend for any convincing call — hang up and ring back on a number you trust, never act on urgency alone, never share one-time codes with a caller who contacted you, and agree a family code word for emergency money requests. Third, watch what Block actually publishes: if Block ever announces an imitation or personal-voice feature, it will come with consent and disclosure rules attached, because the company's own framing — a voice AI that cannot hear distress is unsafe — cuts the other way too: a voice that can pretend to be someone is a safety feature problem, not a party trick. When the safety researchers publish their first findings, cloning and impersonation will be among the most important sections, and we will cover them here.
Sources: Block's official Dessa page (block.xyz/dessa), re-read November 8, 2026, including its demo and GPT-Live-1 notes; DessaAI Daily's verified-facts record on the Dessa team's history and safety-first rollout, October 2026.
Frequently asked questions
Can Dessa AI clone my voice?
Block has announced no voice cloning feature for Dessa, and has never demonstrated Dessa imitating a specific person. In Block's demos Dessa speaks in its own voice and personality. Dessa is also not publicly available, so no cloning tool based on it can be genuine.
Did Dessa clone Joe Rogan's voice?
No — that was the earlier Toronto startup Dessa, in 2019, years before the 2026 model. That synthetic Rogan demonstration was a warning about voice imitation, and the team behind it later became the Block lab that built today's Dessa.
Could Dessa be used to impersonate someone in a scam?
There is no evidence it has been, and Dessa is not public. Impersonation is one of the risks Block's safety-researcher-first rollout is designed to probe before any wider release; no findings have been published yet.
🚀 Get Dessa launch alerts
Dessa is still a research preview. Join the free DessaAI Daily list and we will email you the day access opens — plus the week's biggest Dessa stories, once a week. No spam, no hype.
Form not loading? Subscribe on the signup page →