Block推出Dessa语音AI模型 支持类人化表情交互

由Jack Dorsey旗下的Block公司推出的新型语音AI模型Dessa已开发完成，目前正在邀请安全研究人员进行测试1。该模型在公司研究实验室开发，具备笑声、停顿和哭泣等类人化表情特征1。

Dessa采用全双工架构，能够同时进行说话和听音，实现更自然的语音交互1。与传统语音AI不同，该模型将情感内置于系统本身而非依赖文本处理管道1。Block表示："与当前语音AI交互体验很糟糕。它把你的声音转换成文本，丧失了大量文本无法传达的信息"1。

在硬件配置方面，Dessa可在单块NVIDIA GPU上运行，并能在消费级计算机（如MacBook M4 Max或更高配置）上本地部署1。目前该模型尚未公开发布1。

Block, the company founded by Jack Dorsey, has unveiled Dessa, a new voice artificial intelligence model capable of operating locally on a single NVIDIA GPU 1. The full-duplex model can simultaneously speak and listen, incorporating human-like characteristics such as laughter, pauses, and crying 1.

According to Block, the current landscape of voice AI interactions falls short of user expectations 1. The company stated that existing systems "convert your voice to text, losing a massive amount of information that text cannot convey" 1. Dessa addresses this limitation by embedding emotional nuance directly into the model rather than relying on text-based processing pipelines 1. The system is designed to run on consumer-grade hardware, including computers such as the MacBook M4 Max or higher configurations 1.

The model remains unavailable to the general public at present, as Block is actively inviting security researchers to test and evaluate the technology before any wider release 1. The development of Dessa took place within Block's research laboratory 1.