Imported from xiejunjie524/handdraw-story-video (SKILL.md). Install upstream with npx skills add xiejunjie524/handdraw-story-video. Copyright stays with the author.

Hand-drawn good-deed story

Create a compact cause-and-effect story, then build it with the bundled scripts and HyperFrames.

Defaults

• Deliver 40 seconds, 720×960, 30 fps, H.264 MP4.
• Use 8 distinct beats of about 5 seconds; accept 7–9 beats and 35–45 seconds.
• Use no narration by default. Select licensed BGM for each story instead of fixing one track.
• Generate one 1K low-saturation color mother image per beat. Never retry image generation automatically.
• Derive line art locally from the exact color image with scripts/make_lineart.py; never generate line and color images separately.
• Reveal line art left to right, then fill color left to right from the same origin.
• Render with one worker and never overwrite a prior render.

Workflow

1. Read docs/story-spec.md and write an eight-row beat sheet: event, characters, setting, key prop, action/reaction, two visual details, and framing.
2. Keep every beat visually distinct. Reject repeated mother images, crops, mirrors, recolors, or zooms used to pad runtime.
3. Copy templates/story-template.json to a project story.json and fill it in. Manually wrap captions; do not use HTML line breaks.
4. Generate one complete color mother image for each beat, sequentially, at 1K. Keep broad paper-white negative space and clean separated contours.
5. Run python scripts/make_lineart.py COLOR OUTPUT_LINE for every approved mother image.
6. Run python scripts/build_story.py story.json hyperframes/index.html --check-assets.
7. Install GSAP through npm and copy node_modules/gsap/dist/gsap.min.js to hyperframes/assets/vendor/gsap.min.js.
8. Run HyperFrames checks, inspect the line and color midpoint of every scene, then render with --workers 1.
9. Verify duration, dimensions, audio stream, black frames, text fit, consistent reveal direction, and final color before delivery.

Visual constraints

• Keep the illustrated action mainly in the lower 45%–55% when captions are present.
• Use at most 2–3 people and two key background anchors per frame.
• Avoid dense architecture, crowds, heavy rain hatching, and line-art extraction that becomes a black mass.
• Vary framing: establish place, show a prop/detail interaction, show action/reaction, and end on a changed communal image.
• Keep captions optional, at 1–3 lines and no more than about 18 Chinese characters per line.
• Add in-scene text only when a physical note, sign, phone, label, or receipt exists in the story.

Use the builder rather than rewriting mask and timeline boilerplate.