# 👥 Multi-Person Changer — AI Workflow for ComfyUI

> Transform, swap, and restyle characters across images and video with a 4-stage AI pipeline.

---

## 🎯 What It Does

This workflow is a complete character transformation pipeline that lets you:

- 🧹 Strip a character — remove clothes, accessories, and background (keep the nude body)

- 🔄 Transfer poses & actions — make Person A move like Person B

- 👗 Generate product-ready outfit images — isolate and showcase clothing

- 🧵 Attach outfits to new body — dress Person A in Person B's clothes

- 🎬 Process video frames — detect scenes and apply changes frame-by-frame

- 🔀 Toggle single/multi-image mode — switch between one-off edits and batch video processing

All with a fixed seed cache — reprocess only when you change the settings, not when you re-queue!

---

## 📋 Requirements (Custom Nodes)

| Node Pack | Purpose |

|-----------|---------|

| InsaneGen (included) | Scene detection from video |

| KJNodes | Set/Get key-value store for images, prompts & config |

| ComfyUI-Easy-Use | Background removal (RMBG) |

| VideoHelperSuite | Video loading & info |

| rgthree-comfy | Seed control, node groups, image comparison |

| ComfyLiterals | Text prompt input nodes |

| ComfyUI-Crystools | Boolean switch (multi/single mode) |

| RvTools v2 | Image batch to list conversion |

---

## 🧠 How It Works — The 4-Step Pipeline

The workflow uses KJNodes SetGet nodes to pass data between processing stages without visual spaghetti. Here's the logic:

### Step 1 — 🧹 Strip

> "Remove any outfit, clothes, objects, accessories and background. Draw this character naked with arms at sides."

Takes your input person image, removes the background via RMBG, and strips the character down to a clean base body.

### Step 2 — 🔄 Pose Transfer

> "Change the actions and poses in Image 1 to match those in Image 2"

Takes the stripped character and applies the pose/action from your reference image.

### Step 3 — 👗 Product Outfit

> "Make a product image for her entire outfit."

Processes the reference outfit image into a clean product-style showcase.

### Step 4 — 🧵 Outfit Attachment

> "Attach the outfit in Image 2 to the person in Image 1. High resolution"

Final compositing — dresses the transformed character with the processed outfit at high resolution.

---

## 🎛️ Key Controls

| Parameter | Description |

|-----------|-------------|

| Seed | Fixed seed for reproducible results with caching |

| Megapixels | Output resolution control (default: 1.0 MP) |

| Scene Threshold | Detection sensitivity for video scene splitting |

| Multi Images toggle | Switch between single-image edit and video batch mode |

| Step 1–4 Prompts | Customize each stage of the pipeline |

| Pose reference | Upload any image as the pose target |

---

## 🖼️ Previews

The workflow includes 5 preview nodes so you can monitor every stage:

- PreviewImage (x5) — view Step 1, Step 2, Step 3, and final output in real-time

- Image Comparer (rgthree) — side-by-side before/after comparison

- PreviewAny — inspect intermediate frame data

---

## 💾 Output

- Final result saved via SaveImage to your ComfyUI output folder

- Video mode: scenes auto-detected and processed individually

- All intermediate images stored in KJNodes workspace for inspection

---

## ⚡ Performance Notes

- Caching: The seed field enables ComfyUI's IS_CHANGED caching — run with the same seed and video, and processing is skipped instantly

- Background removal uses RMBG-1.4 (lightweight, runs on most GPUs)

- Video mode uses PySceneDetect for fast scene boundary detection

---

## 🔧 Troubleshooting

- Scene threshold too low? Increase to merge more frames per scene

- Wrong outfit attachment? Tweak the Step 4 prompt for better compositing

- Multi-images not working? Make sure the toggle is ON and a video is loaded

- Need different result? Change the seed — each seed gives a unique variation

---

Made with ❤️ for the ComfyUI community.

Klein9b Extract product

https://civarchive.com/models/2366596/extract-product-image-qwen-and-klein?modelVersionId=2666606

Klein9b High resolution

https://civarchive.com/models/2436859/high-resolution?modelVersionId=2760799

Klein9b Copy pose

https://civarchive.com/models/2380153/copy-pose-qwen-and-klein?modelVersionId=2701726