Jane-xiaoer/paper-collage-ad-codex

Codex skill for complete paper-cut collage ad production, local IndexTTS-2 voice cloning, animation, audio and MP4 QC

What it solves

This project provides a complete workflow for creating paper-collage style advertisements. It automates the process from initial creative ideation, scripting, and storyboarding to the final production of animation, voiceovers, music, and sound effects, resulting in a high-quality MP4 video.

How it works

It is designed as a "skill" for OpenAI Codex, which reads a central SKILL.md file to orchestrate various tools and scripts. The workflow involves:

  • Creative & Scripting: Extracting visual metaphors from product data to create scripts and time-coded storyboards.
  • Visuals: Generating style-locked paper-collage keyframes using brand assets.
  • Animation: Using Gemini Omni (via ChatCut) for rapid motion previews and local tools like FFmpeg or specialized services (Seedance, HyperFrames) for final high-resolution animation.
  • Audio: Generating voiceovers using standard TTS or local zero-shot voice cloning via IndexTTS-2 MLX for Apple Silicon Macs, combined with music and sound effects.
  • Finalization: Compositing all elements into a validated H.264/AAC MP4 file.

Who it’s for

Content creators and advertisers who use OpenAI Codex and want to produce stylized, professional paper-collage animations with a mix of local AI tools and cloud APIs.

Highlights

  • End-to-end pipeline: Covers everything from the first idea to the final MP4 export.
  • Local Voice Cloning: Integrates IndexTTS-2 MLX for high-quality, private, local voice cloning on Mac.
  • Rapid Prototyping: Uses Gemini Omni for quick 3-10 second motion previews to test pacing and humor before final rendering.
  • Modular Tooling: Supports a variety of animation engines (Seedance, Kling, HyperFrames) and audio tools.

Related

  • Project
  • Dispatch
  • Project
  • Project
  • Project