Choose Rich Live
Tech

Block unveils Dessa AI voice model trained entirely on human conversation

Jack Dorsey's Block has unveiled Dessa, a speech-to-speech AI model trained solely on spoken human conversation with zero text that can run locally on an M4 Max MacBook Pro.

Graphic demonstration of Block's Dessa voice AI displaying audio waveforms and emotional reactions during a conversation.
Image: @ai_for_success

Jack Dorsey's Block has introduced Dessa, a full-duplex speech-to-speech artificial intelligence model trained entirely on human conversations without using text data or pretrained models. According to Block, the model is compact enough to run locally on an Apple MacBook Pro equipped with an M4 Max chip or newer.[1][5][6][8][10][11]

Dessa is engineered to engage in conversational audio by listening and speaking at the same time. The model can detect emotional nuances, pick up on sarcasm and sighs, and laugh mid-sentence. Demonstrations shared alongside the launch also showed conversational interactions including fast backchanneling, interruptions, and audible reactions like gasps.[1][2][8][11]

The release attracted immediate attention from technology outlets, with The Verge sharing audio sessions and Gizmodo describing the resulting voices as uncanny and human-ish. However, the lifelike cadence also triggered debate: reporting from Crypto Briefing and CoinDesk noted that while figures like CZ eyed possible applications in children's education, observers warned that such convincing imitation raises ethical concerns over trust, fraud, and impersonation.[3][5][7][9][12]

Key facts

  • Block unveiled Dessa, a full-duplex speech-to-speech AI model trained entirely on human conversation without text data or pretrained models.
  • Block stated that Dessa is capable of running locally on a MacBook Pro with an M4 Max chip or newer.
  • Dessa can listen and speak simultaneously while responding to emotional cues, laughing mid-sentence, and registering sighs and sarcasm.
  • Public reaction was divided between awe at the human-like quality of the voice and concerns that it could facilitate scams and impersonation.
  • CZ indicated interest in potential applications of Block's voice AI model in children's education.

Sources · 11 sources

  1. CB

    Crypto Briefing@Crypto_BriefingPost on X ·

    🚨 NEW: Jack Dorsey's Block introduces Dessa, an AI voice model trained entirely on human conversations without text data or pretrained models. The model can listen and speak simultaneously, recognize emotional cues and run locally on a MacBook Pro with an M4 Max chip. https://t.co/3wjGFmLwa4

    Open source
  2. CO

    Cointelegraph@CointelegraphPost on X ·

    ⚡ JUST IN: Jack Dorsey's Block just built an AI that learned to talk like a baby does, by listening, and it laughs mid-sentence when you're funny. Creepy or cool? https://t.co/TthIo2lWym

    Open source
  3. CO

    CoindeskArticle ·

    Block's new voice AI sounds eerily human, sparking awe and scam fears Reactions split between amazement at how human it sounds and fears it could fuel scams and deception, with CZ eyeing uses in children’s education.

    Open source
  4. AS

    AshutoshShrivastava@ai_for_successPost on X ·

    New edition of AI Compass is live 🧭 (Oct 23) - Moonshot AI launches Kimi K3.1 Pro with 2.8T parameters and 1M context - Block unveils Dessa, a full-duplex speech model trained with zero text Read all the major AI news of the last 24 hours here 👇 https://t.co/hZzCWCbB0o

    Open source
  5. TE

    TechmemeArticle ·

    Block unveils Dessa, a "true speech-to-speech" AI model trained on "nothing but human conversation" that it says runs locally on a MacBook Pro M4 Max or newer (Jay Peters/The Verge) Jay Peters / The Verge : Block unveils Dessa, a “true speech-to-speech” AI model trained on “nothing but human conversation” that it says runs locally on a MacBook Pro M4 Max or newer — You have got to listen to these conversations with Block's new Dessa AI model.

    Open source
  6. TE

    TechCrunch@TechCrunchPost on X ·

    Block said that the voice model can run locally on a MacBook M4 Max https://t.co/Y6sjnpUSpb

    Open source
  7. CB

    Crypto BriefingArticle ·

    Jack Dorsey’s Block unveils Dessa, an AI voice model trained entirely on human conversations Dessa's development highlights the potential for AI to revolutionize human-like interaction, raising ethical concerns about trust and impersonation. The post Jack Dorsey’s Block unveils Dessa, an AI voice model trained entirely on human conversations appeared first on Crypto Briefing .

    Open source
  8. AS

    AshutoshShrivastava@ai_for_successPost on X ·

    Dessa from Block was probably the biggest AI announcement yesterday. Looks like we’ve solved voice AI. Full-duplex, trained entirely on audio with zero text, and runs locally on an M4 Max MacBook Pro ( as per blog) https://t.co/2BGt8xuhSJ

    Open source
  9. GI

    Gizmodo@GizmodoPost on X ·

    Check Out the Many Uncanny, Human-ish Voices From Block's New AI Model https://t.co/QdNp90Clpd

    Open source
  10. PO

    Polymarket@PolymarketPost on X ·

    JUST IN: Jack Dorsey's Block launches “Dessa,” an AI voice model trained entirely on speech that he says is small enough to run on a MacBook.

    Open source
  11. CB

    Coin Bureau@coinbureauPost on X ·

    🚨This is genuinely WILD, you can now talk to an AI like it's a HUMAN. Twitter founder Jack Dorsey's Block just unveiled Dessa, a speech-to-speech model that learned to talk by listening, with no text at all. It listens and talks at the same time, with emotion, laughing mid-sentence, hearing sarcasm and sighs.

    Open source
  12. TV

    The Verge@vergePost on X ·

    You have got to listen to these conversations with Block’s new Dessa AI model. https://t.co/UfAgv45W4c

    Open source