mufeng-bilingual-storybook

name: mufeng-bilingual-storybook description: Create bilingual AI picture books from Markdown source material. Use when Codex is asked to read story markdown, split it into scenes, write child-friendly Chinese and English narration, create image prompts, generate page illustrations with Codex built-in image generation rather than OpenAI Images API, and assemble Chinese/English PDF storybooks from generated images.

Mufeng Bilingual Storybook

Core Rule

Use Codex built-in image generation for artwork. Do not call OpenAI Images API, do not require OPENAI_API_KEY, and do not create one-off image SDK clients.

Use the bundled helper script for deterministic text/PDF work:

python "$CODEX_HOME/skills/mufeng-bilingual-storybook/scripts/mufeng_storybook.py" --help

If CODEX_HOME is unset, use $HOME/.codex.

Workflow

1. Load all Markdown files in the project except build/cache/vendor folders.
2. Choose the page count from context unless the user asks for a fixed count.
   • Short story / compact single episode: 12 pages
   • Medium chapter / several clear beats: 16 pages
   • Formal picture-book chapter: 20 pages
   • Long or complex chapter: 24 pages
3. Write:
   • scenes.parsed.md
   • prompts.generated.md
   • manifest.json
4. Generate one full-page illustration per scene using Codex built-in image generation.
5. Copy final images into <output-dir>/images/.
6. Build both PDFs from the same images:
   • Chinese: storybook.pdf
   • English: storybook_en.pdf
7. Verify page count, image count, and sample rendered pages.

Chinese PDF Font Rule

For Chinese PDFs, do not use Songti.ttc without an explicit font index. On macOS, index 0 resolves to Songti SC Black, which makes captions look too heavy.

Use these explicit Songti weights for Chinese PDF rendering:

• Body captions and footer: /System/Library/Fonts/Supplemental/Songti.ttc, index 3 (Songti SC Light)
• Page titles: /System/Library/Fonts/Supplemental/Songti.ttc, index 6 (Songti SC Regular)

This avoids prior glyph-rendering mistakes where characters such as 着, 将, 起, 径, and 造起 looked blurred or were misread, without making the text visually too bold.

Recommended Commands

Create scene and prompt files only:

python "$HOME/.codex/skills/mufeng-bilingual-storybook/scripts/mufeng_storybook.py" \
  --project-dir . \
  --output-dir build/storybook_auto \
  --scene-count auto \
  --prompts-only

Build PDFs from already-generated images:

python "$HOME/.codex/skills/mufeng-bilingual-storybook/scripts/mufeng_storybook.py" \
  --project-dir . \
  --output-dir build/storybook_auto \
  --scene-count auto \
  --use-existing-images \
  --language both

Run a dry layout check without real images:

python "$HOME/.codex/skills/mufeng-bilingual-storybook/scripts/mufeng_storybook.py" \
  --project-dir . \
  --output-dir build/storybook_auto_dryrun \
  --scene-count auto \
  --dry-run \
  --language both

Force an exact page count when needed:

python "$HOME/.codex/skills/mufeng-bilingual-storybook/scripts/mufeng_storybook.py" \
  --project-dir . \
  --output-dir build/storybook_12 \
  --scene-count 12 \
  --prompts-only

Image Generation Procedure

The helper script intentionally does not call image APIs. For each block in prompts.generated.md:

1. Use the built-in image_gen tool with the prompt text.
2. Save/copy the resulting image into <output-dir>/images/ using the expected filename from manifest.json.
3. Keep generated images free of readable text, labels, captions, borders, and watermarks.
4. After images exist, run the helper with --use-existing-images --language both.

When generating many pages, do not compress distinct scenes into variants of one prompt. Generate one image per scene.

Scene And Translation Guidance

For classic Chinese source text:

• Use --scene-count auto by default, and override only when the user gives an exact page count.
• Use short two-sentence narration per page.
• Keep Chinese narration natural and avoid rare glyphs when previous rendering shows font artifacts.
• Translate for children, not literally. Preserve plot, emotion, and clarity.
• Keep page titles short.
• Put no readable text inside images, even when the source mentions tablets, plaques, or stone inscriptions.

For detailed guidance, read references/workflow.md.