How to read this page. AMBER = simulated previs, computed in your browser from a still and a depth guess (a simplified depth-parallax camera; not model output). GREEN = real model output made once through the Comfy Router API or Comfy Cloud on 2026-10-16: unedited, prompt-only, no camera LoRA, partner models (not open weights, not LTX, except the two LTX-2.3 clips). The clips on this page are web copies, re-encoded smaller and without audio. Credits and request ids are the real ones.
1 Source frame
2 Direct the camera
Pick a still and a move
3 The real clips
Click a clip to play it. Own ratings out of 10 are mine and subjective; credits are measured by balance delta (211 credits per US dollar).
4 Coverage: six real angles of one still
FLUX 3 Image edits of the alley still via the Comfy Router (6 requests, 86.9 credits, unedited). The close-up shows a small garbled, unreadable emblem on the scooter: it is generated, not a real logo.
5 Optional: run it for real with your own key
This page has no backend. If you have a Comfy API key with credits you can render the selected still and move yourself: your browser calls api.comfy.org directly. The key is kept in this tab's memory only (never saved, never put in the URL, never sent anywhere except api.comfy.org) and is gone when you close or reload the tab. Every run costs your credits, is submitted once and never retried, and needs an explicit confirmation. Not tested by anyone but the author, with a few cents of credits. Prefer the repo if you would rather not paste a key into a web page.