Audio#Video generation#Image generation

story-to-handdrawn-video: turn a Chinese story into a hand-drawn page-flip animation

An open-source agent skill turning Chinese stories into 3:4 hand-drawn flipbook MP4s — 327 style assets, hand-written text in the art, Remotion rendering, MIT.

Project facts

GitHub Ecosystem
Repositorygithub.com/gnipbao/story-to-handdrawn-video
License
MIT
Language
HTML
Stars
2,367
Data checked
2026-11-04

Snapshot figures reflect the check date and may change over time.

Colored-pencil diary comic style sample: an elder and a child crossing a stone arch bridge hand in hand

Between a Chinese story and a publishable video sit four jobs: storyboarding, illustration, titling and editing. gnipbao’s open-source story-to-handdrawn-video (MIT, 2,367 stars as of November 4, 2026) hands the whole pipeline to an agent you already have: story text in, a 3:4 portrait hand-drawn flipbook MP4 out — storyboards, image generation, asset processing and Remotion rendering in one pass, with the Chinese hand-written captions drawn directly into the artwork by the image model rather than overlaid in post.

Colored-pencil diary comic style sample: an elder and a child crossing a stone arch bridge hand in hand

Core features

  • Story in, film out: a natural-language agent skill covers storyboarding, illustration, post-processing and Remotion rendering, producing a silent picture-track MP4 that takes voiceover later.
  • 327 style assets: 297 style recipes plus 30 theme palettes across 11 material categories; a built-in 75-style, 45-second page-flip showcase (1080×1440, 30 fps) doubles as a demo.
  • Hand-written text drawn in: Chinese captions are painted into the source images by default, animated layer by layer — text, then pencil sketch, then color — or as a full-page page-flip.
  • Reference images for new stories: the agent reads line work, shapes, palette and texture from a reference image and keeps new storyboards in the same visual language, so series content never drifts.
  • A private style library: name and keep any reference style locally; recall it by name next time.

Typical use cases

  • Kids’ story and picture-book channels: 3:4 portrait fits Douyin’s video accounts and Xiaohongshu, and the flipbook’s handmade feel outlasts photoreal AI footage.
  • Series IP content: lock a character’s look with a reference image and keep the same visual language across daily updates.
  • Style scouting: the online style library filters by material, type or number; all 75 showcase styles use the same bridge-crossing scene, so comparison costs nothing.

Quick start

git clone https://github.com/gnipbao/story-to-handdrawn-video.git

Hand the story text and a style number to an agent with the SKILL.md installed and it renders the MP4. Browse the online style library first (story-handdrawn-style-library.vercel.app) — every asset has a built-in sample image, and the repo ships an offline style-library.html.

Summary

story-to-handdrawn-video fits creators running kids’ stories, picture books or hand-drawn content channels; photoreal styles, live-action texture and long-form narrative are out of scope — the 3:4 hand-drawn flipbook is the one form it makes. Facts and limits: MIT licensed, silent picture track only (add your own BGM; the demo video’s music has attribution terms to follow); output quality tracks your agent’s image tool — the style recipes lower choice cost, not the floor; 2,367 stars since July 2026 with fast iteration, so style numbering will move. For the static sibling of this idea, see handraw-style — numbered prompts for still images — and for the editing stage, video-shotcraft.