AI video examples and recorded workflows
Explore motion, camera direction and scene transitions in public AI video examples, with prompts or methods where available.
A short video may require reference frames, several shots and editing. Check the recorded process: a missing prompt or model setting is unknown, and some cases provide only a reference method.
Folding Venice: The Yellow Envelope
A 30-second continuous cinematic sequence follows a young courier carrying a yellow envelope through a 1940s Venice that folds, rotates, and hangs upside down around her. The author labels the workflow GPT Image 2.5 + Seedance 2.5 and publishes the full character, style, and timecoded video prompt. The prompt requires an image_1 face and identity reference that is not present in the post or visible author replies, so generation and Remix remain disabled.
Model not recorded
Read case & prompt →
Papilio Mechanica: An Interactive Three.js Butterfly
A 28-second screen recording presents an interactive mechanical butterfly in Three.js, moving between assembled and exploded views with light and dark themes. The author shared only the short method instruction shown here, without the full code or parameters. The post calls the model Gemini 4 Pro but says it appeared in Arena under the placeholder gemini-3.7-flash and was later removed, so the model identity and reproducibility remain unverified; generation and Remix are disabled.
Model not recorded
Read case & prompt →
Coastal Downhill Longboarding
A 30-second vertical action video follows a longboarder racing down a winding coastal mountain road above a turquoise bay. The post includes a complete cinematic Prompt calling for three riders, ground-level tracking, aerial views and realistic speed, but does not identify the generation model, so the piece is published as a reference with generation and Remix disabled.