Dessa

★★★★☆4.0 / 5 editorial Free Research preview; public waitlist open, with early access first to a small group of AI safety researchers — no public release date or pricing announced yet.

Full-duplex voice AI trained only on human speech — it listens while it talks.

✓ Editor-researched listing — details reviewed by Grab AI Tool editors; pricing re-checked regularly.

Try it Free → ♡ 851 saves

#voice-ai#speech-to-speech#full-duplex#conversational-ai#research-preview

What is Dessa?

Dessa is a full-duplex speech-to-speech model from Block's small in-house AI research lab, unveiled on October 23, 2026. Unlike most voice AI, which converts speech to text before generating a reply, Dessa takes audio in and gives audio out directly — so it listens while it talks, cuts in at natural moments, backchannels with a quick “yeah” or “got it,” and hears emotion, laughter, sarcasm, and pauses that a text pipeline throws away. Block says it was trained from scratch on nothing but human conversation, with no text and no pretrained models.

Block says Dessa runs locally on a MacBook Pro with an M4 Max chip or better, with inference faster than real time, and has published recorded demo conversations showing it shifting tone, pace, and emotion mid-chat. For now it remains a research preview: a public waitlist is open on Block's Dessa page, but there is no public release date, app, download, or API yet, so anyone curious should treat the demos as a preview of where voice AI is heading rather than a tool they can use today.

Who Dessa Is Best For

Best for: Dessa suits podcasters, voice artists, and audio producers especially well: The most human-feeling voice demos of any speech model released this year; local inference is a real privacy and latency advantage over cloud-only voice AI. It is completely free, so there is no budget decision to make.

Not ideal for: It may not suit you if any of these are dealbreakers: not publicly usable yet — a research preview with early access going first to a small group of AI safety researchers; no benchmarks, model size, release date, or pricing published so far.

Key Features

Full-duplex conversation: listens and speaks at the same time, like a real phone call
Trained only on human speech — no text pipeline and no pretrained models, per Block
Hears and produces emotion, laughter, sarcasm, pauses, and natural interruptions
Runs locally on a MacBook Pro M4 Max or better, with faster-than-real-time inference claimed
Public demo conversations and an open waitlist on Block's official Dessa page

Pros

The most human-feeling voice demos of any speech model released this year
Local inference is a real privacy and latency advantage over cloud-only voice AI
Skipping the text step preserves tone and emotion other voice systems lose

Cons

Not publicly usable yet — a research preview with early access going first to a small group of AI safety researchers
No benchmarks, model size, release date, or pricing published so far
Its strikingly human voice has already raised scam and deception concerns from fraud watchers

Pricing

Model: Free
Research preview; public waitlist open, with early access first to a small group of AI safety researchers — no public release date or pricing announced yet.

Adding Up Your AI Bill?

If you pay for Dessa alongside other AI subscriptions, the monthly total adds up quietly. Tick every tool you pay for in our free AI Tool Cost Calculator to see your real monthly and yearly bill — and spot overlaps you can cut. Still choosing? Start with our guide on how to choose the right AI tool.