What it does

This extension plugs into AUTOMATIC1111’s Stable Diffusion WebUI (and Vladmandic’s fork) to generate those mesmerizing infinite-zoom videos you’ve seen on social media. You provide prompts, it outpaint-expands each frame while shrinking the previous center, then interpolates between steps to create a smooth zoom illusion. The math is simple: Zoom Speed × FPS = frames per outpaint step.

The interesting bit

The cleverness is in the workflow automation, not a new model. It chains img2img outpainting with center-cropping and interpolation—tedious to do by hand, trivial to mess up. The extension hides that bookkeeping behind a single “Generate video” button. There’s also a standalone Google Colab version that runs on free tier, suggesting the authors want this accessible beyond WebUI users.

Key highlights

• Requires an inpainting model for “killer results” (authors’ words, not mine)
• Warns explicitly against mask blur and color correction—common pitfalls that “hit results with the ugly stick”
• Parameters documented in a wiki; UI is a single tab with sliders
• Works on free Colab with a separate Gradio notebook

Caveats

• README is sparse on technical details: no mention of VRAM requirements, generation times, or maximum resolution
• The “How it works” section is more user guide than algorithm explanation—unclear if interpolation is simple crossfade or something smarter
• 670 stars but development activity unclear; contribution guidelines point to a develop branch that may or may not be active

Verdict

Worth a spin if you’re already in the AUTOMATIC1111 ecosystem and want quick zoom videos without scripting. If you’re not using WebUI or need fine-grained control over the interpolation logic, the standalone Colab or manual ComfyUI nodes will likely serve you better.