Comic Generation ( comic)#

Reference Images (Important)#

If you use reference images (image-to-image / series reference / consistency refs):

• Reference images must be public URLs.
• HTTPS is strongly recommended.
• http:// may work but is insecure and can be blocked by some networks.
• Local file paths and data: URLs are not supported by the WeryAI gateway.

Generate educational or narrative comic pages: a knowledge comic, tutorial comic, biography comic, educational comic, or a multi-page comic sequence — structured art, tone, layout, and language decisions included.

This skill turns source material into a sequence of comic pages.

Maintain the art and tone mapping when the bundled generation runtime updates, and re-check recent runtime behavior if a newer version changes page rendering.

Before the first generation run in a new OpenClaw or local instance, run:

bash

npm run ensure-ready -- --project . --workflow comic

This step is mandatory. It reads the doctor report and automatically runs bootstrap when local script dependencies such as pdf-lib are still missing.

If the report shows a missing IMAGE_GEN_API_KEY and the user approves, run npm run setup -- --project . --workflow comic --persist-api-key when the key is already in env, or persist it to .image-skills/comic/.env on the user's behalf, then continue the comic workflow without leaving this skill.

When this skill is first connected, tell the user that the default generation model is Nano Banana 2 ( GEMINI_3_1_FLASH_IMAGE). Also tell them it can be switched later whenever another model fits the task better.

Script:

• scripts/scaffold.ts
• scripts/build-prompts.ts
• scripts/build-batch.ts
• scripts/merge-to-pdf.ts

Safety & Scope#

• Network: This skill calls the WeryAI gateway over HTTPS ( https://api.weryai.com).
• Auth: Uses IMAGE_GEN_API_KEY. The key is never printed. It may be persisted only when you explicitly run npm run setup -- --persist-api-key.
• Reference images: Must be public URLs ( https:// recommended). http:// may work but is insecure. Local file paths and data: URLs are rejected.
• No arbitrary shell: The generation runtime does not execute arbitrary shell commands.
• Files written: Output images and optional local config under .image-skills/comic/ (project) and/or ~/.image-skills/comic/ (home).

Use Cases#

• knowledge comics
• tutorial comics
• biography comics
• multi-page narrative image sequences

Not a good fit for:

• a single cover image
• a dense one-page infographic
• a RedNote card series

Core Dimensions#

1. art
2. tone
3. layout
4. aspect

See:

• references/dimensions.md
• references/character-template.md
• references/storyboard-template.md
• references/prompt-template.md

Commands#

ScriptPurposescripts/scaffold.tsInitialize storyboard, characters, and page promptsscripts/build-prompts.tsRegenerate page prompts from storyboard.md``scripts/build-batch.tsGenerate batch.json from page promptsscripts/merge-to-pdf.tsMerge page images into a single PDFscripts/package-delivery.mjsPrepare delivery bundle with manifest and previewsnpm run generateGenerate page images./scripts/vendor/compression-runtime/scripts/main.tsCompress output for delivery

Workflow#

Step 1: Initialize Working Files#

Create the working directory:

bash

${BUN_X} {baseDir}/scripts/scaffold.ts \
  --output-dir comic/topic-slug \
  --title "Comic Title" \
  --topic "Topic summary" \
  --art ligne-claire \
  --tone neutral \
  --layout standard \
  --aspect 3:4 \
  --scope standard \
  --lang en \
  --panels 3 \
  --pages 4

This creates:

• analysis.md
• storyboard.md
• characters/characters.md
• prompts/00-cover.md
• prompts/01-page.md
• ...

Step 2: Understand the Content#

Extract:

• the main story arc or knowledge arc
• the characters
• the key scenes
• the target aspect ratio
• the expected total content scope and number of pages
• the target panel density per page
• the user's language, especially if dialogue or captions appear on the page

Step 3: Choose art, tone, and layout#

Default priorities:

• art: ligne-claire
• tone: neutral
• layout: standard
• aspect: 3:4
• scope: standard
• panels: 3

Recommended rules:

• educational explanation -> ligne-claire + neutral
• high-energy storytelling -> manga + action
• warm storytelling -> ligne-claire + warm
• ink or wuxia-like themes -> ink-brush + dramatic

Step 4: Map to the Bundled Runtime#

The bundled image runtime does not directly understand comic page grammar, so:

• map art to --style
• write tone, layout, camera language, and panel logic into the prompt body
• create one prompt per page
• prefer batch execution for multi-page comics

Recommended mapping:

comic artruntime --style``ligne-claire``flat-illustration``manga``manga``realistic``photoreal``ink-brush``ink-brush``chalk``chalk

Step 5: Refine storyboard.md, Then Build Prompts#

Save at least:

• analysis.md
• storyboard.md
• characters/characters.md
• characters/characters.png
• prompts/00-cover.md
• prompts/01-page.md

When the storyboard is ready, generate prompt files automatically:

bash

${BUN_X} {baseDir}/scripts/build-prompts.ts \
  --storyboard comic/topic-slug/storyboard.md \
  --output-dir comic/topic-slug/prompts

Each page prompt should clearly describe:

• the page goal
• the characters and scene
• the cross-page continuity anchors
• the panel structure
• the intended panel count
• the target language for any text on the page

Character consistency rules:

• default to the strongest continuity strategy: generate one canonical characters/characters.png first, then reuse that same reference for every page
• create characters/characters.md before generating multi-page comics
• generate characters/characters.png as a reference sheet before page generation
• if the chosen model supports --ref, pass the same characters/characters.png to every page
• if the chosen model does not support --ref, copy the key character descriptions into every page prompt
• include page-level continuity anchors in storyboard.md so outfit, props, location, and scene progression stay aligned from page to page
• each page prompt should include previous-page context and next-page hook so the sequence reads like one comic, not isolated posters

Recommended sequence:

1. finish storyboard.md
2. finish characters/characters.md
3. generate characters/characters.png
4. make sure every page in storyboard.md includes Continuity Anchors
5. batch-generate all comic pages

Step 6: Build batch.json#

When page prompts are ready, generate a batch file:

bash

${BUN_X} {baseDir}/scripts/build-batch.ts \
  --prompts comic/topic-slug/prompts \
  --storyboard comic/topic-slug/storyboard.md \
  --ref comic/topic-slug/characters/characters.png \
  --output comic/topic-slug/batch.json \
  --images-dir comic/topic-slug \
  --model "$M" \
  --jobs 4

The script:

• reads 00-cover.md, 01-page.md, 02-page.md, and so on
• maps comic art from storyboard.md into runtime --style when possible
• defaults to the same shared character reference sheet for every page from storyboard.md when --ref is not explicitly overridden

Default best-performing strategy:

1. generate characters/characters.png first
2. keep that sheet as the canonical look for faces, outfits, props, and palette
3. reuse the same reference for every cover/page task
4. only fall back to text-only continuity when the chosen model cannot reliably use references

Step 7: Run Generation#

Generate the character sheet first:

bash

${BUN_X} {baseDir}/npm run generate \
  --promptfiles characters/characters.md \
  --style manga \
  --image comic/topic-slug/characters/characters.png \
  --ar 4:3 \
  -m "$M"

Single-page example:

bash

${BUN_X} {baseDir}/npm run generate \
  --promptfiles prompts/02-page.md \
  --style manga \
  --ref comic/topic-slug/characters/characters.png \
  --image comic/topic-slug/02-page.png \
  --ar 3:4 \
  -m "$M"

Batch example:

bash

${BUN_X} {baseDir}/npm run generate \
  --batchfile comic/topic-slug/batch.json \
  --jobs 4

Step 8: Merge Pages into PDF#

After page images are ready:

If this skill was distributed as a slim package without node_modules, run npm run bootstrap from this skill directory once before this step so the local PDF dependency is installed. If you forget, merge-to-pdf.ts now prints that exact recovery hint.

bash

${BUN_X} {baseDir}/scripts/merge-to-pdf.ts comic/topic-slug

To choose a specific output path:

bash

${BUN_X} {baseDir}/scripts/merge-to-pdf.ts comic/topic-slug -o comic/topic-slug/final.pdf

The script merges page files in the order 00-cover-*, 01-page-*, 02-page-*, and so on, across png, jpg, and jpeg.

Step 9: Prepare Delivery Files#

When you need a cleaner handoff than a raw table of URLs, prepare a delivery bundle:

bash

${BUN_X} {baseDir}/scripts/package-delivery.mjs comic/topic-slug

This creates:

• delivery/preview.md with inline preview entries for small batches
• delivery/manifest.json with ordered file metadata
• delivery/pages/ with copied page images
• delivery/<topic-slug>-delivery.zip automatically when the page count is large enough

Delivery rules:

• Few pages (≤ 5): show each page image directly in order (cover → pages). Do not just list file paths.
• Many pages (> 5): show the first 2–3 pages as preview, then provide the PDF and/or zip bundle.
• Always offer the merged PDF as the primary deliverable.
• Ask if any pages need changes before finalizing.
• Auto-compress: once confirmed, run the bundled compression runtime on the output directory before packaging.

bash

${BUN_X} {baseDir}/./scripts/vendor/compression-runtime/scripts/main.ts comic/topic-slug/ -r -f webp -q 80

Internal checklist (for agent): art / tone / layout / aspect, page count, character reference used, model, PDF generated, compression done.

Output Convention#

Suggested output directory:

text

comic/<topic-slug>/

Suggested minimum files:

• analysis.md
• storyboard.md
• characters/characters.md
• characters/characters.png
• prompts/00-cover.md
• prompts/NN-page.md
• batch.json
• 00-cover.png
• NN-page.png
• <topic-slug>.pdf

Re-run Behavior#

• scaffold.ts on an existing directory overwrites storyboard.md, characters/characters.md, and all prompt files. Back up before re-scaffolding.
• build-prompts.ts overwrites prompt files in prompts/ based on the current storyboard.md.
• build-batch.ts overwrites batch.json.
• Re-running the bundled generator with --batchfile regenerates all pages; keep good pages by removing their entries from batch.json first.
• merge-to-pdf.ts overwrites the existing PDF.

Definition of Done#

• storyboard.md, characters/characters.md, and per-page prompt files exist in the output directory.
• All pages are generated and shown to the user in reading order.
• A merged PDF is produced via merge-to-pdf.ts.
• Art style, page count, and model are stated in the delivery summary.
• A compressed webp set is produced for delivery.

Iteration#

When the user wants changes after seeing generated comic pages:

• Art style mismatch ("wrong style", "switch to ink brush") → change art in storyboard.md, update --style mapping, re-generate affected pages. Ask if all pages or specific ones.
• Character inconsistency ("characters look different across pages") → regenerate characters/characters.png with more explicit descriptions, then re-generate pages using the updated reference sheet.
• Panel layout issues ("bad panels", "this page is too crowded") → revise the specific page in storyboard.md, rebuild its prompt, re-generate only that page.
• Story flow issues ("wrong order", "need an extra page") → revise storyboard.md to add/reorder pages, rebuild prompts for changed pages, re-generate.
• Single page redo → re-generate only that page with --promptfiles prompts/NN-page.md. Keep other pages.

After any page changes, re-run merge-to-pdf.ts to update the PDF.

Current Scope#

This version of comic focuses on:

• comic-page workflow
• character consistency
• page-level prompt organization
• single-gateway image execution