Turn a MiniMax H3 reference-to-video multishot generation into a finished manga or comic page, all inside ComfyUI.
Design the page first: each panel becomes one hard-cut shot of the video, and each shot is turned back into its panel.
Komikku Page & Shots: fullscreen page layout editor (presets, free, slanted and polygon panels, knife cuts). Click a panel to write its shot; the node builds the H3 multishot prompt with timestamps.
MiniMax H3 ref2vid generates the multishot video from up to 4 character references.
Scene detection picks one frame per shot.
FLUX 3 Klein reframes each frame to its panel's exact aspect ratio.
Optional MOSS-Image hi-res pass up to 5 MP.
Komikku Page Compose: fullscreen editor to move and zoom images inside panels and add lettering (speech, thought, shout, whisper, caption, SFX...), RGB/RGBA overlays, a favorites library and custom fonts.
Optional Enhance with CLIP: rewrites shots and suggests dialogue with a local text encoder (MOSS Qwen3.5-9B).
Included
The main workflow plus 4 genre examples (Superheroes, Detective, Manga, Dark). Load your own character into Picture 1 before running. The Manga example uses a screentone LoRA (manga_tone_v2, trigger "manga-tone rendering").
Install
Custom nodes: github.com/quinteroac/ComfyUI-Minimax-H3-Komikku
Other custom nodes needed (listed in the workflow note): ZNGB-SceneDetect, VideoHelperSuite, Pixaroma, rgthree, comfyui-obvpm.
MIT license. Thanks to Pixaroma, Manga Maker Studio, obvpm, rgthree and the Anima workflows for inspiration.
Description
First release: Komikku Page & Shots, Komikku Page Compose and Komikku Fit Frame nodes, main workflow (1:1 video) and 4 genre examples.



