Claude Opus 5.5 Video List Publishes 475 Prompts
The awesome-opus5-5-videos repository has published 475 Claude Opus 5.5 prompt files associated with generated videos, along with a data/videos.json index. Its practical consequence for developers is direct: the document
The awesome-opus5-5-videos repository has published 475 Claude Opus 5.5 prompt files associated with generated videos, along with a data/videos.json index. Its practical consequence for developers is direct: the documented workflow is to copy a prompt into Claude Code, the Claude app or another agent running Opus 5.5, request a single HTML file, and record the rendered page.
The README highlights 100 entriesβone for each distinct promptβsplit into 58 motion graphics, 16 explainers, 14 3D scenes and 12 games or interactive projects. The latest listed update, dated September 29, 2026, says 86 videos were added, mostly in the motion-graphics and explainer categories.
From creator post to browser remake
Every entry links to its creatorβs original post. The repository also directs readers to Skillry, where all 475 originals can be viewed beside live remakes, according to the project page.
The project says every listed video was made by asking Claude Opus 5.5 to write animation as code, using technologies such as HTML, Canvas, SVG and Three.js. Individual entries carry more specific implementation labels, including GLSL, WebGL, GSAP, CSS, Web Audio, particles and playable interactions, as shown in the repository index.
The inputs are not uniformly complete. Some creators shared only part of a prompt, while others published a description of the result rather than the prompt itself; the corresponding files identify those limitations, according to the README.
Analysis: a pattern library, not a benchmark
Analysis: The uneven inputs make this collection more credible as a prompt-pattern archive than as a controlled test of model performance. Because the project acknowledges partial and missing prompts, a difference between an original and its remake cannot establish that the published words alone caused the result. The side-by-side view supports visual inspection, but not strict reproducibility claims.
The repository carries an MIT license, while its README separately says each video and prompt belongs to its creator and invites creators to request changes or removal through an issue. Analysis: That leaves the scope of reuse less explicit than the copying workflow. Developers can inspect and test the published inputs, but the page does not reconcile its repository license with the stated ownership of each underlying prompt and video.
Originally published by Dev.to AI. Aggregated on AIWithGhost for educational purposes β full credit and traffic to the original publisher.