docvideoer AGENTS.md

AGENTS.md — docvideoer

This project supports Codex agent workflows for turning documents into narrated Remotion videos.

Project Overview

docvideoer converts a document URL or pasted text into an explainer video. The agent coordinates document extraction, storyboard generation, TTS synthesis, Remotion project setup, rendering, and final output reporting.

External Skill Dependency

Remotion-specific implementation depends on Remotion's official remotion-best-practices skill. When creating, debugging, or changing the Remotion project, scene components, timeline, animation, audio sync, captions, assets, or rendering flow, use $remotion-best-practices if it is installed. If it is missing, tell the user to install it with:

npx skills add https://github.com/remotion-dev/skills --skill remotion-best-practices

Trigger Intent

When the user says /docvideo, "把这个文档转成视频", "根据这篇文章生成视频", or otherwise asks to create a video from a URL/text document, run the document to video workflow below.

Output Layout

Create each video project under:

docvideo-output/<project-name>/

Expected files:

source.md                 # fetched or pasted source content
storyboard.json           # structured storyboard
tts-batch.json            # optional batch input for TTS
audio/                    # scene audio and optional VTT subtitles
video/                    # generated Remotion project
output.mp4                # rendered video
output-with-subtitles.mp4 # optional subtitle-burned video

Use a short, filesystem-safe project name derived from the document title or topic.

Workflow

1. Get document content.

   • For URLs, run:

     node scripts/fetch-doc.mjs <url> --out docvideo-output/<project-name>/source.md
     

   • For pasted text, write it to source.md.

2. Generate storyboard.json.

   • Extract the document's core points.
   • Use concise scenes, each focused on one idea.
   • Estimate Chinese narration at about 4 characters per second.
   • Use these scene types: title, content, bullet, image, end.