Analysis updated 2026-05-18
Turn a short spoken line into a stylized explainer video for social media.
Generate paper collage style marketing clips without hiring a designer.
Add a synced voiceover to an auto generated visual metaphor.
| megatroll222/vox-collage-broll | ahammadmejbah/awesome-llm-datasets | akseolabs-seo/fable-soul | |
|---|---|---|---|
| Stars | 105 | 105 | 105 |
| Language | — | — | Python |
| Setup difficulty | moderate | easy | easy |
| Complexity | 2/5 | 1/5 | 2/5 |
| Audience | vibe coder | researcher | developer |
Figures from each repo's GitHub metadata at analysis time.
Needs Claude Code plus ffmpeg installed locally, and a MaxFusion account or your own Google AI Studio and ElevenLabs API keys.
This project is a skill for Claude Code, the AI coding assistant, that turns a single spoken sentence into a short animated video. You give it a line, for example "A snowball only needs the first push, the hill does the rest," and it builds a five second, vertical video in a paper collage art style: flat bold colors with black and white cut out photo pieces that look like they were snipped from magazines and assembled by hand. The process happens in three stages, and you approve the result at each one before moving to the next. First the skill suggests a visual metaphor for your sentence. Once you approve that idea, it generates a single finished still image in the collage style. Once you approve the still, it animates the whole scene assembling itself from an empty colored background, piece by piece, like stop motion. This staged approval means you never pay for a video generation on an idea you have not already signed off on. The skill also checks its own work. It reviews contact sheets and individual frames to confirm the first frame is truly empty, the last frame matches the approved still, and no fake or garbled lettering has crept into the image, and it has a set of fixes ready for the failure patterns that come up in practice. There is an optional final step that adds a spoken voiceover: you pick a voice from your library, the line is generated as audio, trimmed to fit the four to five second clip, and mixed in without touching the video itself. To use it you need Claude Code with ffmpeg installed on your machine, since frame preparation and the final video mixing happen locally, plus a connection to MaxFusion, a service that handles the actual image, video, and speech generation. The skill can also be adapted to use other providers directly, such as Google's image and video tools and ElevenLabs for voice, if you already hold your own API keys. This is an English language adaptation of an original project called gbro collage broll, credited to its creator pyang5166, who designed the visual style, the three step approval flow, and the repair process this version reuses. It is released under the MIT license.
A Claude Code skill that turns one spoken sentence into a short paper collage style explainer video, with staged approvals and self QA.
MIT: free to use, modify, and share, including commercially, as long as you keep the copyright notice.
Setup difficulty is rated moderate, with roughly 30min to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
Verify against the repo before relying on details.