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

2026-04-24

AI Summary

Current generative video models rely heavily on prompt-based interaction, lacking controllability and transparency in their internal mechanisms, which limits artists’ creative expression. This work introduces a plugin system within the DayDream Scope ecosystem that enables, for the first time, fine-grained real-time manipulation of video diffusion Transformers. The system supports independent intervention in self-attention, cross-attention, and feedforward modules at the levels of diffusion timesteps, network layers, prompts, and individual neurons. By introducing the novel concept of “material affinity” to enhance interpretability, the approach functions as a new kind of AI art probe, allowing users to intuitively perceive model dynamics and explore creative possibilities beyond the default representational space.

Abstract

Generative video models have achieved remarkable visual fidelity, yet their prompt-only interface offers thin creative agency and obscures the model's material process from the artists working with it. We present Real-Time AttentionBender, a tool that extends the practice of network bending across the full depth of the video diffusion transformer (DiT) and brings it into live, interactive generation. Built as a plugin within the DayDream Scope ecosystem and wrapping open-source real-time Wan pipelines, the tool exposes self-attention, cross-attention, and the feed-forward network as independently manipulable surfaces, with targeting down to individual diffusion steps, DiT layers, prompt tokens, and hidden neurons. The immediacy of live manipulation affords what we call"material intimacy"with the model: a responsive, near-mechanistic feel for how specific layers and neurons shape generated video. We position the tool as simultaneously an XAIxArts probe into transformer internals and an expressive instrument for discovering aesthetics outside the model's default representational space.

Problem

• generative video models
• creative agency
• model interpretability
• interactive generation
• transformer internals

Innovation

• Real-Time AttentionBender
• video diffusion transformer
• interactive network bending
• material intimacy
• XAIxArts