Block Launches Full-Duplex AI Voice Model Dessa for Local MacBook Operation

icon币界网
Share
AI summary iconSummary
Block, the company behind Jack Dorsey’s latest venture, has released Dessa, a full-duplex AI voice model capable of running locally on high-end MacBooks such as the M4 Max. Trained on real human voice data, Dessa can replicate emotional cues like laughter and sobbing, and it can listen and speak simultaneously. Block states that the model was trained on dialogue data to enhance natural interaction. While public testing is not yet available, the company is collaborating with security experts to evaluate potential risks. On-chain data indicates rising interest in voice AI, with the Fear & Greed Index reflecting increased market sentiment around innovation.
CoinDesk reports:

IT Home, October 25: Block, the company founded by Twitter co-founder Jack Dorsey, has launched a new voice model called Dessa, which supports local execution and can simulate human-like emotional expressions such as laughter, pauses, and sobbing. According to the official description, Dessa was trained exclusively on human voice data using just one rack of NVIDIA GPUs.

Dessa is a full-duplex model, like OpenAI’s GPT-Live released earlier this year, capable of speaking while simultaneously listening to user speech. While mainstream voice models typically rely on large language models to access knowledge, Block states that Dessa’s knowledge is derived directly from the conversational data used in its training.

Block stated that current voice AI experiences are often stiff and unnatural, primarily because the generation pipeline first converts content to text and then uses that text to drive speech synthesis. Many models, such as ElevenLabs v4 and Fish Audio Drama 3, rely on voice labels to control emotion. However, this company founded by Dorsey claims that emotional expression is natively built into the model, though it remains unclear whether its underlying system still depends on some form of labeling to regulate emotion.

Block wrote in the blog: "The current voice AI interaction experience is truly poor. It first converts your speech into text, directly losing information that cannot be conveyed through large amounts of text. Even the world's most advanced AI cannot detect your emotions. We find this unreasonable—not only is the experience subpar, but it also poses potential risks."

The research team released multiple dialogue samples with Dessa, an agent capable of switching voice tone, speech rate, and emotion. For example, in one dialogue, the AI says it is tired from work, and its vocal delivery simultaneously conveys a sense of fatigue.

In another example, the AI speaks about the sense of comfort the person brings, with an emotional tone that is almost tearful. In another demonstration, a Block employee plays a game with the voice assistant.

According to official documentation, this model can run locally on consumer-grade computers, with a minimum requirement of a MacBook equipped with an M4 Max or higher. However, Dessa is not currently open for public trial; Block is extending testing invitations to security researchers to focus on risk scenarios involving voice agents deceiving humans.

According to IT Home, Block’s core business is financial services, providing solutions for small and medium-sized enterprises; its subsidiary Square is already using voice AI to handle order processing. After developing this on-device voice model, Block is expected to deploy more voice-based products for merchants and general users at a lower cost.

Disclaimer: The information on this page may have been obtained from third parties and does not necessarily reflect the views or opinions of KuCoin. This content is provided for general informational purposes only, without any representation or warranty of any kind, nor shall it be construed as financial or investment advice. KuCoin shall not be liable for any errors or omissions, or for any outcomes resulting from the use of this information. Investments in digital assets can be risky. Please carefully evaluate the risks of a product and your risk tolerance based on your own financial circumstances. For more information, please refer to our Terms of Use and Risk Disclosure.