Repository Index

We propose Prism, a dynamic sparse attention framework for natively training joint video-audio generation models at 2K.

• diffusion-models
• high-resolution
• sparse-attention
• video-generation

Official implementation of "AReA: Adaptive Receptive-Field Attention for Efficient High-Resolution Diffusion Transformers"

• diffusion-models
• efficient-inference
• high-resolution
• image-generation