Block, the fintech company behind Square and Cash App, has unveiled a new AI model named Dessa that marks a significant shift in how emotional intelligence is processed in artificial systems. Unlike traditional text-based models, Dessa is trained exclusively on human speech, enabling it to detect nuances like emotion, sarcasm, and pauses that are often lost in transcription-based approaches.
Speech-First Approach for Emotional Intelligence
The model’s architecture is built to analyze audio directly, bypassing the need to convert speech into text first. This allows Dessa to capture subtle vocal cues—such as pitch, tone, and rhythm—that are essential for understanding the emotional context of a conversation. The system is designed to operate in real time and is currently optimized to run locally on Apple’s latest MacBook Pro M4 Max, ensuring privacy and low latency.
Implications for AI and Human Interaction
While Dessa is not yet available to the public, its development signals a growing interest in AI systems that mirror human communication more accurately. As voice interfaces become more prevalent in customer service, mental health support, and educational tools, models like Dessa could redefine how AI interprets and responds to human emotion. The ability to process speech without relying on text conversion may also improve accessibility for users who struggle with traditional text-based systems.
What’s Next for Dessa?
Although Block has not announced a public release date, the company’s move toward audio-first AI models could influence the broader tech industry. As voice technology continues to evolve, we may see more companies investing in models that understand not just what people say, but how they say it. Dessa could be a key step toward more emotionally intelligent AI systems that truly connect with users on a human level.