comic-maker
github.com
- Free plan
- Windows
- Linux
- Self-hosted
What comic-maker is
comic-maker turns a story idea into a lettered comic using models and software that run locally. An Ollama chat model drafts the title, characters, pages, and panel details as JSON, which users can edit before continuing. ComfyUI with z-image turbo checkpoints generates an image for each panel; the project says users can adapt its generation function for another model. A PIL compositor assembles pages with one to nine panels, diagonal gutters, speech bubbles with tails, Impact sound effects, and a title on the first page. Styles include sunday-strip, manga, noir, and euro-album, or users can describe another style. Per-character seed families and repeated visual descriptions in prompts are intended to help keep characters consistent. Script writing, panel generation, and page assembly can be run as separate stages. The tool exports PNG pages and a CBZ archive. It is MIT-licensed and free, with no cloud service, credits, or accounts required. The early release is actively developing, and its API, prompts, and layouts may change. It supports Windows, Linux, and self-hosting.
Who it suits
comic-maker suits comic creators who want to turn a story idea into pages with locally run models. Users need Python 3.10 or later, Ollama with a chat model, and ComfyUI with z-image turbo checkpoints.
Good to know, and what to watch
- Runs locally without cloud, credits, or accounts
- Editable JSON separates script and generation stages
- Supports four built-in styles and custom descriptions
- Exports PNG pages and CBZ archives
- MIT licensed and free
- Requires Python, Ollama, and ComfyUI prerequisites
- Speech bubbles currently anchor at the top center
- API, prompts, and layouts may change during development
The verdict
comic-maker offers a local workflow from story draft to assembled comic pages, with several style and export options. It is an early release, and its current speech-bubble placement is limited to top-center anchoring.
What the maker publishes
- Character consistency
- The project uses a per-character seed family and repeats the character’s visual description in each panel prompt to help maintain consistency. source
- Development status
- The project describes itself as an actively developing early release and says its API, prompts, and layouts may continue to change. source
- Export
- The compositor produces PNG pages and a CBZ comic archive. source
- Features
- Page assembly supports 1–9 panels per page, diagonal gutters, wrapped speech bubbles with tails, Impact sound effects, and a title on page one. source
- Image generation
- ComfyUI with z-image turbo generates one image per panel, and the README says users can edit the generation function to use another model. source
- Integrations
- The documented workflow uses Ollama for a chat model and ComfyUI with z-image turbo checkpoints for panel images. source
- Intended audience
- The README says the project was made with local models on one PC. source
- Known limitation
- Speech bubbles currently anchor at the top center; character-aware placement is listed on the roadmap. source
- Known limitations
- The roadmap says speech bubbles currently anchor top-center and lists character-position-aware placement and multi-character bubble tails as future work. source
- License
- The repository states that comic-maker is MIT licensed. source
- Local use
- The README says the project uses local models on one PC and requires no cloud, credits, or accounts. source
- Page composition
- The PIL compositor supports 1–9 panels per page, diagonal gutters, speech bubbles with tails, Impact sound effects, a title on page one, and CBZ export. source
- Privacy
- The README describes the tool as fully local, with no cloud, credits, or accounts. source
- Purpose
- Comic-maker turns a story idea into a lettered comic using a local language model, a local image model, and a compositor. source
- Requirements
- The quickstart lists Python 3.10 or later with Pillow, Ollama with a chat model, and ComfyUI with z-image turbo checkpoints as prerequisites. source
- Script generation
- A local Ollama model creates the title, characters, pages, and panel details, and users can edit the resulting JSON by hand. source
- Styles
- Built-in styles are sunday-strip, manga, noir, and euro-album, and users can also provide a style description. source
- Support
- The README invites users to follow the repository and submit issues and ideas. source
- Support and feedback
- The README invites users to follow the repository and submit issues and ideas. source
- Workflow
- Script, panel generation, and assembly can run as separate stages, with the script JSON editable between stages. source
Checked against the maker's pages on 8 Oct 2026.