jane-xiaoer/paper-collage-ad-codex — explained in plain English
Analysis updated 2026-05-18
Generate a short paper cutout style advertisement video for a product from a text prompt.
Produce a script, storyboard, and keyframes for a stop motion style ad automatically.
Add AI narration to a video using a locally cloned, permission cleared voice.
Assemble animation, music, and sound effects into a quality checked final MP4.
| jane-xiaoer/paper-collage-ad-codex | heigeai/heige-codex-skin-studio | yyx990803/vue-svelte-size-analysis | |
|---|---|---|---|
| Stars | 312 | 314 | 309 |
| Language | JavaScript | JavaScript | JavaScript |
| Last pushed | — | — | 2022-12-06 |
| Maintenance | — | — | Dormant |
| Setup difficulty | moderate | — | moderate |
| Complexity | 3/5 | 2/5 | 2/5 |
| Audience | vibe coder | vibe coder | developer |
Figures from each repo's GitHub metadata at analysis time.
Basic keyframe and animation use needs no API key, voice cloning and some effects need extra local setup or third party accounts.
This project is a skill for OpenAI Codex that walks through making a paper cutout style collage advertisement video, start to finish. It is installed into a Codex project or into a user's global Codex skills folder, and once installed you can ask Codex in plain language, such as making a 45 second paper cutout ad for a product, and Codex reads the included instructions to carry out the steps. The workflow covers the whole production pipeline: pulling a visual theme out of product materials, writing a script and dialogue with a shot by shot storyboard, generating paper cutout style keyframe images that stay consistent with real brand assets, animating those frames using tools like Seedance, HyperFrames, layered PNGs, or FFmpeg, adding narration through regular text to speech or through a local voice cloning tool, and finishing with music, paper rustling sound effects, and a final MP4 that passes a quality check before delivery. For voice cloning it relies on a separate local project called mlx-indextts, built for Apple Silicon Macs, which can clone a voice from a short reference clip. The README is explicit that no one's voice or voice model ships with the repository, that cloning should only be done with a voice the user owns or has clear permission to use, and that any delivered video should disclose the narration was AI generated. Reference audio and generated voice files are meant to stay local and are excluded from version control by the project's gitignore template. Basic use, meaning static keyframes, layered animation, and final assembly, needs no API keys at all. Optional services like Seedance, Jimeng, MiniMax, and ElevenLabs each require the user's own account and credentials, read from environment variables rather than stored in the repo. On macOS, setup involves installing ffmpeg and Node with Homebrew and running a dependency check script. The skill's own original content is released under the MIT License, while any third party models, fonts, music, or assets it can call out to follow their own separate licenses.
A Codex skill that automates making a paper cutout style collage advertisement video, from script and keyframes through animation, voice, and a final quality checked MP4.
Mainly JavaScript. The stack also includes JavaScript, Node.js, FFmpeg.
The skill's own content can be used freely for any purpose, including commercial use, as long as you keep the copyright notice, third party models and assets have their own separate licenses.
Setup difficulty is rated moderate, with roughly 1h+ to a first successful run.
Mainly vibe coder.
This repo across BitVibe Labs
double-check against the repo, no cap.