DiffusionBrowser: Interactive Diffusion Previews via Multi-Branch Decoders

TL;DR

DiffusionBrowser introduces an interactive, multi-modal decoder framework for diffusion models that enables real-time video previews and user-guided generation, enhancing transparency and control during the synthesis process.

Contribution

It presents a lightweight, model-agnostic decoder enabling interactive, real-time video previews and control during diffusion-based generation.

Findings

Generates multi-modal previews at over 4x real-time speed

Allows interactive guidance via stochasticity reinjection and modal steering

Provides insights into the denoising process and scene composition

Abstract

Video diffusion models have revolutionized generative video synthesis, but they are imprecise, slow, and can be opaque during generation -- keeping users in the dark for a prolonged period. In this work, we propose DiffusionBrowser, a model-agnostic, lightweight decoder framework that allows users to interactively generate previews at any point (timestep or transformer block) during the denoising process. Our model can generate multi-modal preview representations that include RGB and scene intrinsics at more than 4$\times$ real-time speed (less than 1 second for a 4-second video) that convey consistent appearance and motion to the final video. With the trained decoder, we show that it is possible to interactively guide the generation at intermediate noise steps via stochasticity reinjection and modal steering, unlocking a new control capability. Moreover, we systematically probe the model using the learned decoders, revealing how scene, object, and other details are composed and assembled during the otherwise black-box denoising process.