New: AI API Cost Calculator — compare GPT, Claude, Gemini, Grok & Llama pricing live
Voice AI · Explainer

Block Dessa Explained: Full-Duplex Voice AI With No Text Pipeline

By ToolsDig Research · Updated October 25, 2026 · How we research

Some links in this article are affiliate links (ElevenLabs and Murf AI) — if you sign up through one we may earn a commission at no cost to you. The main subject, Dessa, is not publicly available and cannot be bought; the disclosure applies to the alternatives section, and it never buys a better verdict.

You can't try Dessa. Block announced its speech-to-speech model on October 23 and the demos sound strikingly human — pauses, laughter, the little "yeah, got it" noises people make without thinking. But there is no API, no pricing, no download and no benchmark sheet: access is a waitlist plus early access for a small group of AI safety researchers. We went through the announcement and launch coverage to separate what Dessa actually is from what the demos merely suggest — and to map what full-duplex changes for call centres, creators and app builders if it works.

🔍 How we put this together

We reviewed Block's Dessa announcement (October 23, 2026), TechCrunch's launch coverage, and independent demo impressions reported by The Verge. Every performance and hardware claim about Dessa is Block's own and is labelled as such; Block has published no parameter count, no API details and no benchmark results, and we have not used the model — nobody outside Block's early-access circle publicly has.
🏆 Verdict: Dessa is a direction, not a product. Speech-native, full-duplex conversation is very likely where voice AI is heading, and Block's speech-only training approach is the boldest version of it yet. But today there is nothing to buy, integrate or benchmark. If you ship voice now, pipeline tools like ElevenLabs (agents and TTS) and Murf (studio voiceover) are the practical answer; treat Dessa as a design brief for 2027, not a roadmap item for this quarter.

The pipeline you're using now — and what it throws away

Almost every voice AI you've talked to is three separate systems bolted together: speech-to-text transcribes what you said, a language model reads the transcript and writes a reply, and text-to-speech reads the reply out loud. It works. It also deletes most of the conversation. Your sigh, your hesitation, the sarcasm in your voice, you talking over the bot because it's heading the wrong way — none of that survives transcription. Text carries words; it doesn't carry how they were said.

Block's argument in its announcement is blunt: the most powerful AIs in the world can't hear how you feel, and that makes voice interaction both worse and, in their framing, potentially dangerous. Anyone who has shouted "representative" at a phone tree knows the feeling.

What Dessa actually is

Per Block's announcement and the launch coverage:

The lab lead, Ragavan Thurairatnam, called it a "GPT-3 moment for speech." That is a big claim from an interested party, and right now it is unfalsifiable either way: no parameter count, no published evaluations, no independent testing. The Verge listened to the demos and found them mostly natural but occasionally awkward — in one, the model mistook the speaker for someone named "Ronaldo."

Full-duplex vs. what you can actually buy

Dessa (Block)GPT-Live (OpenAI)Typical voice-agent stack
ArchitectureSingle speech-to-speech modelFull-duplex modelSTT + LLM + TTS pipeline
Text in the middle?No — audio in, audio outNoYes — everything passes through a transcript
Interruption / overlapNative, per BlockNativeBolted on via voice-activity detection and barge-in handling
Knowledge sourceIts conversation trainingA frontier model lineageWhatever LLM you plug in — swappable
AvailabilityResearch preview, waitlistShipping productBuild it today from APIs
BenchmarksNone publishedVendor-reportedYou measure it yourself, in production

The honest read: full-duplex is not a Block invention — OpenAI's GPT-Live got there as a product first. Dessa's bet is different: that removing text entirely produces conversation that feels more human than any pipeline, however good the components get.

Advertisement

What this changes if it works

Call centres. This is where full-duplex bites first. Real callers interrupt, talk over prompts, cry, get sarcastic. A model that hears all of that — and doesn't need a neat transcript to respond — handles the messy share of calls that current bots fumble into "let me transfer you." But here is the catch no demo answers: Dessa's knowledge comes from conversation, not from your help centre, your CRM, or last Tuesday's pricing change. A voice that sounds human but can't look anything up is a very charming dead end. The winning support product is probably full-duplex ears and mouth on top of a retrievable brain — which is, ironically, a pipeline again, just a better one.

Creators. Scripted text-to-speech is fine for narration. It is bad at dialogue — two voices reacting to each other, timing jokes, leaving space. A conversational model that generates natural back-and-forth could do podcast banter, character scenes and interview-style content that today needs two humans or heavy editing. We're not there yet — there is no product — but this is the use case to watch.

App builders. Voice stops being a chat window with a microphone button. If the model can listen while it speaks, interface design changes: no push-to-talk, no waiting for the beep, no "sorry, I didn't catch that" loops. The constraint moves from transcription accuracy to conversation management — turn-taking, interruption policy, and knowing when the agent should just stop talking and listen.

The catches, collected in one place

No benchmarks. Zero. Every capability claim traces back to Block's demos. Knowledge ceiling: conversation-trained knowledge is frozen at training time and shallow by design — fine for chat, dangerous for facts. The manipulation problem is real: a model that hears emotion can also perform it — synthetic warmth, synthetic urgency. Block giving safety researchers first access suggests they know this; expect disclosure rules to follow the capability. Demos are not products: a waitlist, a rack-sized training run and a handful of recordings are a research announcement, not a release.

What you can use today

Until Dessa (or something like it) ships, voice work still runs on pipelines. Two tools cover most of the actual jobs:

ElevenLabs — the default pick for both studio-quality text-to-speech and real-time conversational agents. The voices are the strongest we've found for narration and character work, and their Agents platform handles the STT/LLM/TTS plumbing for phone and web voice agents. The catch: it is the pipeline architecture — emotion is steered through controls and tags rather than learned, turn-taking still needs tuning, and per-minute agent costs add up fast at call-centre volume. Check current pricing before you commit; it changes often.

Murf AI — a studio voiceover tool, and we mean that as a boundary, not a compliment or a criticism. If your job is e-learning narration, explainer videos or ads read in a consistent voice, Murf's editor is faster and simpler than ElevenLabs. The catch: it is not a conversational agent. Nothing you build in Murf will listen to anyone. Buy it for narration, never as a step toward a Dessa-style product.

And if you just want to feel full-duplex today, GPT-Live inside ChatGPT is the shipping example — no purchase needed, and ten minutes of conversation teaches more about the interaction model than any article.

The bottom line: today, Dessa changes what you should design for, not what you should buy. Ship on pipelines, design for interruption, and re-check the waitlist when Block publishes an API, pricing or a single independent benchmark.

FAQ

Can I use Block's Dessa today?

No. As of October 2026, Dessa is a research preview. Block runs a public waitlist and is giving early access to a small group of AI safety researchers. There is no public API, no pricing, no downloadable weights and no published benchmark results — everything public so far is Block's own demo recordings.

What does full-duplex voice AI mean?

It means the model listens and speaks at the same time, the way people do in conversation. You can interrupt it, talk over it, laugh or add a quick 'yeah' while it is talking, and it can react in real time. Most voice AI today is half-duplex: it waits for you to finish, transcribes your words, generates a text reply and reads it aloud.

How is Dessa different from other voice AI?

Dessa was trained from scratch on human conversation only — no text and no pretrained language model underneath — so it takes audio in and produces audio out with no transcription step. Most voice systems chain speech-to-text, a language model and text-to-speech, which loses tone, timing and emotion. According to Block, Dessa's knowledge comes from its conversational training data rather than a bolted-on language model.

What hardware does Dessa run on?

Block says Dessa runs locally on a MacBook Pro with an M4 Max chip or better, with inference faster than real time. It was trained on a single NVIDIA GB200 NVL72 rack. These are Block's own claims and have not been independently verified.

What voice AI tools can I use instead of Dessa today?

For conversational voice agents and high-quality text-to-speech that you can actually buy today, ElevenLabs is the most complete option, covering both studio voices and real-time agents. For studio voiceover work such as e-learning, explainers and ads, Murf AI offers a simpler editor. If you want to experience full-duplex conversation now, OpenAI's GPT-Live inside ChatGPT is a shipping example.