Difforum By Chillithebillis Turns Sound Into Moving Video Art

    
        By vramkickedin    
     | 
    
            July 7, 2026 at 6:19 pm        
    
     | 
    
        2 min read

Difforum is a set of custom nodes for ComfyUI that creates keyframe animations using math expressions, camera moves, and audio reactivity. It updates the classic Deforum workflow to work with modern image models like SDXL, Flux, and SD3.5. The system offers three render paths to generate videos with depth-aware warping or prompt travel.

Developer Chillithebillis built this project to fix compatibility issues found in the original Deforum code. The older version refused to run on Python 3.12 and was restricted to older image models. This rebuild removes those limitations while maintaining the familiar keyframe syntax and adding new animation tools.

Animation nodes for comfyui users

Key Features

Model agnostic sampler works with SDXL.
Audio reactivity turns sound into curves.
Camera presets provide intuitive zoom controls.
Live sampler displays realtime generation preview.
Three render paths handle video workflows.

This tool suits creators who want to generate music visuals or abstract videos directly inside their existing ComfyUI setup. Users can connect audio files to drive video effects, making it useful for live performances or edited content. Anyone needing precise camera control combined with modern image generation will find the node interface helpful.

Project notes and technical details

The project avoids using unsafe code evaluation by relying on a secure evaluator for math expressions. It includes 16 test suites to ensure the warp, audio, and color effects function properly across different hardware. Users running Apple Silicon devices should use fp16 or GGUF model versions because fp8 checkpoints do not work on Mac hardware.