pyang5166/gbro-collage-broll

What it solves

It automates the creation of high-end, editorial-style "halftone paper-collage" B-roll animations from short voiceover scripts (approximately 5 seconds). It prevents wasted API costs by implementing a strict human-in-the-loop approval process before generating expensive video content.

How it works

The project uses a "three-gate approval" workflow to ensure aesthetic quality and cost efficiency:

1. Gate 1 (Metaphor Confirmation): The system proposes a visual metaphor, key objects, background colors, and assembly order for approval without generating any media.
2. Gate 2 (Still Frame Confirmation): Once the metaphor is approved, it generates a colored collage still frame and a contact sheet for review.
3. Gate 3 (Video Generation): After the still frame is approved, it uses gemini-omni-flash-preview to generate a first-and-last-frame assembly animation with a stop-motion feel.

Who it’s for

Content creators and editors who need stylized, professional B-roll clips to overlay on voiceovers without manually designing complex collage animations.

Highlights

• Stop-motion Aesthetic: Elements slide in and assemble from an empty field rather than using simple fades or zooms.
• Cost-Control Workflow: Mandatory three-stage approval process to minimize API spend on incorrect generations.
• Specific Output Format: Delivers 9:16, 5-second, 720x1280, 24fps silent MP4 files.
• Automated Setup: Includes self-checking scripts to manage dependencies like ffmpeg and the Google GenAI SDK.