# H3 Look Sheets v1.1 — character + outfit reference sheets with MiniMax H3

Put a person photo and a few outfit photos in. Get a contact sheet of that person, wearing that exact look, from every angle.

H3 Look Sheets is a ComfyUI custom node pack and workflow built on MiniMax H3's multi-reference mode (ref2va). It describes your reference photos with a vision model, writes the shot-by-shot H3 prompt for you, renders one take that moves through every angle you asked for, picks the best frames, and lays them out as a clean look sheet.

Useful for character consistency, fashion/lookbook sheets, LoRA dataset prep, and costume references.

---

## ✨ What's new in v1.1 — multi-image references

- Several outfit photos at once (up to 8). Split the look across images: front/back of the top, front/back of the bottom, shoes, and up to 4 accessories (glasses, bag, jewelry…). Tell the node which image shows what, and each photo is described only for its own piece. No more duplicate shoes because two photos showed them.

- Several person photos. Front, back, close-up… All are used as identity references. Hair and hairstyle come only from the person photos and stay identical across all shots.

- New Image Aggregator node. Routes every person and outfit image into H3's 9 reference slots, in the right order, automatically.

- Automatic gender detection from the person description (can be overridden).

- More robust descriptions. Refusals, empty or "nothing visible" answers are detected and retried with a new seed. An image that still fails is skipped instead of polluting the prompt.

- Updated example workflow: 2-stage sampling + upscale.

---

## 🔧 How it works

1. Load the person photo(s) and the outfit photo(s).

2. Describe Reference: a Qwen3-VL vision model writes one sentence per image (person: face, hair, body; outfit: garments and accessories only).

3. Shot Config ×1–15: pick angle (front, 3/4, profile, back…), framing (full body → extreme close-up) and expression for each shot.

4. Look Sheet Prompt - Custom Shots writes the full H3 ref2va prompt, with every <Picture N> tag matching the wired images.

5. MiniMax H3 renders the take.

6. Select Frames picks the keepers: content clustering, optional vision-assisted diversity, sharpness fallback.

7. Datasheet Settings lays them out as one contact sheet (columns, tile width, padding).

## 🧩 Nodes

| Node | Role |

|---|---|

| Look Sheet Prompt - Custom Shots (H3) | Builds the ref2va prompt from 1–15 shots |

| Shot Config (H3 Look Sheet) | Angle / framing / expression for one shot |

| Describe Reference (H3 Look Sheet) | Vision description of up to 8 photos, with retry + fallback |

| Image Aggregator (H3 Look Sheet) | Routes person + outfit images into H3's 9 reference slots |

| Select Frames (H3 Look Sheet) | Picks the best, most varied frames from the render |

| Datasheet Settings (H3 Look Sheet) | Lays frames out as a contact sheet |

---

## 📦 Requirements

- MiniMax H3 ref2va MiniMaxH3ReferenceToVideo, core ComfyUI node). The example workflow uses the int8 ref2va model + the 4-step turbo LoRA.

- Qwen3-VL as the vision CLIP (the example uses qwen3vl_4b_fp8_scaled) for Describe Reference and, optionally, Select Frames tier 2.

- No extra Python dependencies. The example workflow uses a few common utility nodes; install any missing ones with ComfyUI Manager.

## 💾 Install

cd ComfyUI/custom_nodes

git clone https://github.com/shisa84/ComfyUI-H3LookSheets

Restart ComfyUI. The nodes appear under H3LookSheets. Drag the workflow from example_workflows/ onto the canvas.

## 💡 Tips

- Keep your references consistent: same person, same hair across all person photos. Inconsistent inputs give inconsistent sheets.

- Crop the head off outfit photos worn by a different model, so their face or hair doesn't leak into your character.

- Assign every outfit image to a slot (top / bottom / shoes / accessory_N). An unassigned close-up (e.g. earrings) gets described as "nothing visible".

- Low VRAM? Favor medium and close-up shots over wide ones. The 2-stage workflow renders small first, then upscales.

- Pick angle and framing that make sense together: back + close-up has no face for the expression to show on.

---

🔗 GitHub: https://github.com/shisa84/ComfyUI-H3LookSheets

📺 Demo videos: https://www.youtube.com/playlist?list=PLR0q7a2fnl5w