Oscilloscope Diffusion reinterprets existing video with prompts, LoRAs and editable timelines
Chuka444 · reddit · 2026-10-04
Developer Uisato Studio unveiled Oscilloscope Diffusion, a novel diffusion-based approach to reinterpreting existing video: the movement and form of the source become the starting point, while textures, materials and visual language are rewritten — e.g. turning audio-reactive geometries into origami, architecture or a Renaissance painting.
Built on the author's TouchDesigner audio-reactive systems, the demo uses material from the Oscilloscopes tool collection (now v1.2), though any video source works. Users pick the source, describe the treatment, and steer fidelity via prompts, curated LoRAs and editable timelines. It's live at Uisato Studio and will be open-sourced soon.
Related event: Oscilloscope Diffusion Reimagines Video Visuals via Diffusion Models(4 posts)→
More from Multimodal
- Looped-DiT: 260M looped model beats 6.5x larger ones with 4.9x less inference compute — burny_tech · 2026-10-04
- Augie's AI Image Browser indexes ComfyUI metadata and offers leak-free exports for selling — spanktastic0x · 2026-10-04
- Music cover generated on 8GB VRAM: YuE2 + LTX-2 render in 17 minutes — big-boss_97 · 2026-10-04
- Image-to-image character translation on Runway turns host into an 'Alternative Late Show' — c_valenzuelab · 2026-10-04
- Immunologist generates 2-minute immunology history video entirely in code with Claude — DeryaTR_ · 2026-10-04
- Seedance 2.5's crisp details expose Minimax H3's missing promised 2K update — Cequejedisestvrai · 2026-10-04