chengyi-ai/native-subtitle-quote-image

What it solves

This project automates the creation of high-quality, long-form "quote images" from video content. It solves the problem of manually capturing frames and aligning subtitles by providing a structured way to turn video clips into visually consistent, shareable image collages, either by preserving original burned-in subtitles or by overlaying verified script text.

How it works

The tool operates in two distinct modes:

• Native Mode: It identifies and crops existing burned-in subtitles directly from the video frames, preserving the original visual style without re-rendering or OCR.
• Script Mode: It allows users to provide a JSON file containing verified dialogue and timestamps, which the tool then overlays onto selected video frames using custom layout settings.

The project is designed to be used as an "Agent Skill" (compatible with tools like Claude Code or Codex), allowing users to trigger the process with natural language commands. It handles frame selection, cropping, layout composition, and quality checks.

Who it’s for

It is for content creators, researchers, or anyone who needs to extract meaningful quotes or dialogue sequences from videos for social media, documentation, or presentations. It is particularly useful for those who want to maintain professional visual standards without manual editing.

Highlights

• Dual Subtitle Handling: Clearly separates native burned-in subtitles from后期 (post-production) script overlays to prevent confusion.
• Agent-Ready: Designed to be integrated into AI agent workflows, allowing for one-sentence natural language commands to process videos.
• Flexible Layouts: Supports various aspect ratios and compact layouts, with options for adding visual separators and speaker prefixes.
• Quality-Focused: Includes built-in checks for frame quality, subtitle alignment, and duplicate detection to ensure the final output is clean and readable.