Skip to main content
GameDev.net gamedev.net
2 sources covering this story

Real-Time AttentionBender: Granular Interactive Network Bending of Video Diffusion Transformers

A new research tool lets artists poke at video diffusion transformers in real time, not just via prompts. AttentionBender exposes self-attention, cross-attention, feed-forward blocks, layers, tokens, and even hidden neurons during live generation. For technical artists and graphics programmers, the interesting part is the shift from black-box prompting to interactive model steering.

First reported 3 months, 1 week ago generative-video diffusion-models transformers attention
Want to discuss what this means for developers?
Open discussion
PRIMARY SOURCE
arXiv cs.GR arXiv cs.GR

Real-Time AttentionBender: Granular Interactive Network Bending of Video Diffusion Transformers

3 months, 1 week ago Read source
arXiv cs.GR arXiv cs.GR 1% match

AI Level of Detail: Distance-Aware ML Model Precision Selection for Real-Time Human Motion Prediction in Games

3 months, 1 week ago Read source