Yongxiang Li

Ripple achieves real-time audio-video generation at 28 FPS while maintaining coherent long-form content, a breakthrough for streaming applications.

Reciprocal cross-stream addressing in Dual Attention Residuals leads to significant improvements in Transformer performance without sacrificing depth-wise diversity.

Transfer learning from a large, pre-trained speech synthesis model unlocks high-quality Tibetan TTS, even with limited Tibetan-specific data.

Open-source audio reasoning just leveled up: Audio-Cogito rivals closed-source models on complex audio tasks.