Workflow Name: Add a Character Model to the Scene
【Workflow Introduction】
This is a workflow designed to add a character model to a scene. There are still many aspects that need improvement, and suggestions are welcome.
After opening this workflow, first upload two images in the lower-left corner of the interface: one is a scene image, and the other is a character image. Draw a mask on the scene image to determine the approximate position of the character. Next, process the image mask, using SegmentAnythingUltra V2 multiple times during this process for segmentation. Then, handle the prompts using the Runninghub Captioner and Midjourney Style Prompter nodes. If you want to use it locally, you can replace these with other reverse-prompt nodes. Moving on to the first sampler, start by using the kolors model for local inpainting. Personally, I prefer using the kolors model for local inpainting, but if you have other better options, feel free to replace it. Then, use the sdxl model for local refinement, followed by the flux model for inpainting. After that, replace background elements and refine the image, optimize the face, and finally use SUPIR for overall high-definition upscaling.

【Usage Scenario】
This workflow can be used to add character models to a scene. Starting from uploading the scene image and character image, it involves a series of steps such as drawing masks, segmentation, and prompt processing. In this way, it skillfully integrates the character naturally into the scene. It effectively eliminates the cumbersome process of hiring models, bringing great convenience to work while significantly improving work efficiency, allowing creators to focus more on their creative ideas.

【Key Nodes】
SegmentAnythingUltra V2, Runninghub Captioner, Midjourney Style Prompter

【Model Versions】
FLUX
Model Name: flux1 dev fp8 Comfy Org.safetensors
kolors
Model Name: Kolors Inpainting.safetensors
SDXL
Model Name: realvisxlV50 v50Bakedvae.safetensors

【LoRA Model】
Model Name: xlab flux Realism2 lora.safetensors

【ControlNet Application】
None

【Custom Sampler (Advanced)】
Sampling Method: euler
Scheduler: simple
Denoising: 0.7