Oscilloscope Diffusion Reimagines Video Visuals via Diffusion Models
Creator uisato (Uisato Studio) released "Oscilloscope Diffusion," a new technique that uses diffusion models to intervene in existing videos, and stated it will soon be open-sourced.
Confirmed
- Takes an original video's motion and forms as a starting point, reinterpreting its textures, materials, and visual language rather than generating from scratch
- Particularly suited to abstract, structure-driven material (such as audio-reactive geometric systems built in TouchDesigner)—the demo's source footage is exactly this kind
- Can transform audio-driven geometric graphics into styles like origami, architecture, or Renaissance oil paintings
- uisato explicitly stated in the original post that the method will be open-sourced soon
Why it matters
- Shifts the focus of generative video from "generating content" to "reinterpreting existing material," offering creative coding communities like TouchDesigner a new path for combining procedural visuals with diffusion models
- Once open-sourced, it should lower the barrier to reproduction and spur more experiments in material and style transfer
2026-10-03 ~ 2026-10-04 · 4 related posts
Primary sources
- Oscilloscope Diffusion Reinterprets Abstract Video Through Controlled Diffusion, Going Open Source — Chuka444 · 2026-10-03
- Oscilloscope Diffusion reinterprets existing video with prompts, LoRAs and editable timelines — Chuka444 · 2026-10-04