Stable Diffusion does video, one flickering frame at a time

What it does

Mov2mov is a plugin for AUTOMATIC1111’s Stable Diffusion WebUI that ingests a video, runs each frame through the image generator, and repackages the results into a new video. It is essentially glue code that turns a text-to-image model into a frame-by-frame video pipeline. The README is light on technical specifics—there is no detail on codec support, resolution handling, or processing speed—but the core loop is straightforward: deconstruct, diffuse, reconstruct.

The interesting bit

The beta Video Editing tab uses keyframe compositing—selecting or auto-generating anchor frames and backpropagating their tags—to dampen the frame-to-frame inconsistency that makes AI video flicker. That temporal coherence trick is the only thing preventing the output from looking like a slideshow in a windstorm.

Key highlights

• Runs inside the familiar AUTOMATIC1111 WebUI as an extension
• Processes video frames individually through Stable Diffusion
• Keyframe compositing (beta) to reduce flicker between frames
• Auto-generate or manually select keyframes for the compositing pass
• Windows-only for the video editing features; other platforms can disable the tab

Caveats

• The beta video editing features are Windows-only
• README is sparse on technical specifics: no mention of codecs, resolution limits, or how the keyframe backpropagation actually works
• Prominent legal disclaimer shifts all copyright and liability to the user

Verdict

Worth a look if you already live in the AUTOMATIC1111 ecosystem and want to experiment with stylized video without leaving the UI. If you need professional video-to-video pipelines, this is too thin and too platform-restricted to be your main tool.