ComicForge

A tiny, scriptable comic-page engine. Characters are a base SVG + stackable variant overlays (faces, arms, …); a comic page is a declarative YAML spec; output is SVG / PNG / PDF. Built so an LLM (or you) can author pages as plain text — see the authoring skill skills/comicforge/SKILL.md for the full authoring contract and skills/comicforge/reference.md for a deeper reference. Working on the engine instead? See CLAUDE.md.

Start a project

ComicForge ships no art and a project needs no Python — a project is just YAML specs + SVG / pixel art that the CLI points at. Install the engine globally and scaffold one:

uv tool install comicforge         # puts `cmf` on your PATH (no venv needed)
cmf init my-comic                  # scaffold characters/ scenes/ pixel/ pages/ + skill
cd my-comic && cmf render pages/hello.yaml

See docs/starting-a-project.md for the data-only setup, version pinning, and when to make it a real Python project.

Working on the engine

uv sync                            # create .venv and install deps

Use

# render a comic page
cmf render examples/pes/pages/slepice.yaml -o slepice.png
cmf render examples/pes/pages/slepice.yaml -o slepice.pdf

# render a standalone scene illustration (one background, no comic grid)
cmf scene examples/pes/pages/dvur-scene.yaml -o dvur.png

# render individual panels at low res for quick review
cmf panel examples/pes/pages/slepice.yaml -o panel.png --row 0 --col 0
cmf panel examples/pes/pages/slepice.yaml -o panels/ --all --scale 0.5

# print the machine-readable contracts (characters / scenes, slots, variants)
cmf characters --library examples/pes/characters
cmf scenes --scenes examples/pes/scenes

Or from Python:

from comicforge import render_spec
render_spec("examples/pes/pages/slepice.yaml", "out.pdf")

How it fits together

ComicForge is a content-free engine plus self-contained projects. There is no shared asset library: every project owns all of its own art. A real downstream project just does pip install comicforge and brings its own characters — none of the example art comes with the engine.

A project is one directory with four asset folders:

examples/pes/
  characters/tom/             flat: base.svg + <slot>-<variant>.svg + character.yaml
  characters/bara/            posed: shared faces + character.yaml + poses/{sit,walk}/
  scenes/{dvur,pokoj}/        base.svg + <slot>-<variant>.svg + scene.yaml
  pixel/{heart,sun,...}.yaml  grid + palette
  pages/
    slepice.yaml              2×2 comic — Tom and Bára watch chickens
    kosticka.yaml             Strip with scenes, pixel art, multiple bubble types
    dvur-scene.yaml           Standalone illustration (use `comicforge scene`)

comicforge/library.py         loads a character, stacks base+overlays, places it
comicforge/scene.py           loads a scene, stacks overlays, scales it to cover a box
comicforge/bubbles.py         speech / thought / shout bubbles with word-wrap + tail
comicforge/pixelart.py        grid+palette → SVG sprite
comicforge/render.py          row/panel pages + standalone scenes → SVG → PNG/PDF
comicforge/cli.py             `render`, `scene`, `panel`, `characters`, `scenes` commands

Asset directories

Each spec declares where its assets live via top-level keys (paths are relative to the spec file's directory). For a spec under pages/, the sibling asset folders are one level up:

library:    "../characters"   # character art
scenes_dir: "../scenes"       # scene backgrounds
pixel_dir:  "../pixel"        # pixel-art sprites

CLI flags (--library, --scenes, --pixel-dir) override spec keys and are treated as relative to the current working directory.

See skills/comicforge/reference.md for the full path resolution rule, how to add characters/scenes/sprites, and the complete CLI reference.

Design choices

• Base + overlays in one shared local canvas — "posing" is just choosing which overlay SVGs to stack. No rig math; each pose/expression is a small SVG you can edit by hand.
• Declarative spec — the whole agent surface is YAML + the characters JSON.
• SVG all the way down, rasterized only at the end — crisp vector PDF for print.
• Content-free engine — comicforge/ ships no art. Each project owns all its own assets, so the engine is a dependency, not a content bundle.

Roadmap ideas

• A FastAPI + HTMX preview server (live re-render on spec edit).
• An MCP server exposing list_characters / render_page as tools.
• Part-anchored slots (separate eyes/mouth/limb pivots) for finer posing.
• Free-form panel shapes and a gutter/layout DSL.
• AI-image panels (drop a generated raster into a panel like pixel art).