Research at Mazed

YAP-1

Most Realistic Voice Model

State-of-the-art realism. Ultra-low latency.

YAP-1 is an AI voice model built for speed and realism. We'll be sharing benchmarks and results soon—followed by an open-source release.

Beyond voice, we're researching ways to optimize MLLMs for performant, real-time conversational environments—pushing the boundaries of applied multimodal AI.

Expected Release: Q1 2025

Interested in exchanging ideas?

[email protected]