H3 Image Character Sheet: one reference in, full character sheet out

Upload a picture of your character (or nothing at all), pick a few settings, and press Run. MiniMax H3 Image draws the whole character sheet in a single image, so every view shows exactly the same character.

What you get

- Top row (full body): front, left profile, right profile, back, plus an optional 3/4 view

- Bottom row (face close-ups): front, left profile, right profile, smiling

Four-legged animals get wide side views, front and back views and three head close-ups. Vehicles get side, 3/4, front, rear and top-down views plus a headlight close-up.

The sheet copies the art style of your reference picture: a photo stays a photo, anime stays anime, watercolour stays watercolour, and 3D stays 3D. Markings, fur, hair, clothing and body proportions stay the same across every view, because every view comes from the same render.

How to use it

1. Upload your character picture(s) in the Reference Images box. Extra pictures should be more angles of the same character. Upload nothing and the workflow designs a brand-new character instead.

2. (Optional) Describe the character in character_notes on the Settings panel. This helps with details the picture doesn't show, like the back of a costume, and you can also ask for an art style here. With nothing uploaded, your notes describe the new character.

3. Set the options on the Settings panel:

- character_type: Human, Creature (two legs), Four-legged animal / creature, or Vehicle

- three_quarter_view: adds a 3/4 view (vehicles get a low-angle "hero" 3/4 shot)

- aspect_ratio: the shape of the sheet, 3:2 by default. Use a landscape shape (3:2, 4:3 or 16:9).

- megapixels: the render size, 16 MP by default. There's no upper limit.

- final_size: the size of the saved sheet:

- Native (full render size)

- 2.3 MP (~1920×1280)

- 4 MP (~2496×1664)

- 6 MP (~3072×2048, default)

- 9 MP (~3744×2496)

- 12 MP (~4352×2896)

- 16 MP (~5024×3360)

4. Press Run.

The render size and the final size are set separately. You can render at high resolution for detail and still save a sheet of a sensible size. The final size only ever shrinks the render, never enlarges it, and Native keeps the full render.

Why it's easy to use

- Just upload, settings and Run. There are no Pose Studio or image-comparison parts.

- All the options are on one Settings panel, built from native ComfyUI nodes. Everything else is tucked inside it. Right-click it and choose Open Subgraph to edit the prompts, the sampler settings or the model file names.

- The layout reads left to right: Step 1 Upload, Step 2 Settings, Step 3 your character sheet.

Requirements

- ComfyUI v0.45.0 or newer. The MiniMax H3 Image Conditioning node isn't in older versions.

- Custom nodes (all installable through ComfyUI-Manager):

- ComfyUI-Pixaroma (the info note and run timer)

- ComfyUI-Fantastic-MiniMaxH3-PromptBuilder, which provides the reference picture loader

- Models:

- diffusion_models/minimax_h3_image_pruned_int8_convrot.safetensors (~21 GB), from [Comfy-Org/MiniMax-H3-Image](https://huggingface.co/Comfy-Org/MiniMax-H3-Image)

- text_encoders/qwen3vl_32b_minimax_h3_int8_convrot.safetensors (~27 GB), from [Comfy-Org/MiniMax-H3](https://huggingface.co/Comfy-Org/MiniMax-H3)

- vae/minimax_h3_image_vae_fp16.safetensors (~1.2 GB), from [Comfy-Org/MiniMax-H3-Image](https://huggingface.co/Comfy-Org/MiniMax-H3-Image)

Tips

- Speed: on my RTX 5090, a 16 MP sheet takes about 30-40 seconds, 4 MP about 20 seconds and 24 MP about 40-50 seconds.

- Drafts: set megapixels to 4 for quick tests, then raise it for final art.

- Image content: the reference should show one character. Use the matching character type for animals and vehicles.

- No reference: with nothing uploaded and no notes, the seed picks a random idea from a list for each character type. Want something specific, like a hybrid creature? Describe it in the notes; the notes always win.

- Left and right: "left" and "right" mean the direction the character faces in the image.

- Reroll: the Seed is set to randomize, so every Run gives a new sheet. If a sheet has wide empty bands at the sides, just run it again.

Credits

- Built on the Comfy-Org MiniMax H3 Image workflow, using the MiniMax H3 Image models.

- Inspired by foxydits H3 Image workflow https://civarchive.com/models/3045966/h3-image-w-pose-studio-other-advanced-features-foxfuressence

What’s new in v2.0

• Character Type dropdown: a native Custom Combo in its own pane under the uploader, with Character Notes beside it. It switches all 8 prompts to one of four sets. Each set is an editable column inside Engine › Shot Prompts.

• Four-legged and vehicle subjects: the body views render landscape automatically. Vehicles get front ¾, rear ¾, top-down and a detail close-up in the bottom row instead of faces. Landscape panels leave white bands above and below on a 3:2 sheet; that’s expected.

• Design first when nothing is uploaded: the workflow detects an empty upload and renders one design image, which then becomes Picture 1 for all 8 shots. With an upload, that extra render is skipped.

• Native nodes only: the new logic adds no custom node packs. It uses CreateList/GetItemFromList, StringFormat + StringCompare, a lazy If/Else Switch and RegexReplace.

What's new in v3.0

Most of these changes came from your comments and messages. Thank you!

- The whole sheet is ONE render. v2 made 8 separate images and stitched them together, so the face and outfit could drift slightly between panels. v3 generates the entire sheet in a single image, so every view shows exactly the same character. It's also much faster: about 30-40 seconds for a 16 MP sheet on an RTX 5090, where v2 took about 3 minutes at a lower resolution.

- Your art style is kept. v2 tended to turn everything photoreal. v3 copies the style of your picture: anime stays anime, watercolour stays watercolour, Pixar-style 3D stays 3D, a photo stays a photo.

- New: 3/4 view toggle. It adds a three-quarter view (head and full body, turned 45°) to the top row. For vehicles it adds a low-angle "hero" 3/4 shot instead.

- No wasted white space. The views fill the whole canvas with only thin gutters between them, so every pixel goes to the character.

- One simple settings panel. All options sit on a single node, built only from native ComfyUI nodes. The workflow reads left to right: Step 1 Upload → Step 2 Settings → Step 3 Result.

- Render at any size. Megapixels are no longer capped at 16.

- Final size dropdown. Render big for detail, then save a smaller file: 2.3 / 4 / 6 / 9 / 12 / 16 MP, or Native to keep the full render. It only ever shrinks, never enlarges.

- Better random characters. With nothing uploaded and no notes, the seed picks one idea from an editable list per type, such as a German shepherd, a Holstein cow, an anthropomorphic raccoon or a 1960s sports car in racing green. No more cat-dog hybrids.

- One fewer custom node pack. rgthree-comfy is no longer needed.