The Rail Wrecked video Archive is a flat, public, indexable catalog of fully short video clips. Every clip on every page is media, produced by a text-to-video or image-to-video model, with zero involvement from any real human subject. No clip is a. No clip uses scraped or filmed footage. Every frame originates from a model.
The archive is organized into 29 pages, 50 clips per page, sorted alphabetically by filename. Pages do not curate or hide outputs, the goal is to publish the full breadth of video output, including the strong examples and the unusual ones, so visitors interested in the actual state of video in 2026 can browse a working dataset rather than a marketing reel.
Each page below lists 50 clips. Click any page to browse its contents. Hover any video card to autoplay. Click a card to open the underlying file directly.