Introduction

Q: What is the purpose of this workflow?

A: This workflow is primarily used for further image blending using the Stable Diffusion webui when the fusion of the character into the background image using Photoshop is not sufficient. This workflow can replace the manual image blending process. However, the premise is that the character has already been composited into the background, not generated from a white background image. The logic of these two workflows is different.

Q: How to use this workflow?

A: Step 1: You need to find the corresponding model pose on the material website and then blend it into the picture using Photoshop. Step 2: Directly import the image into the workflow to get a high-quality model image that avoids copyright issues.

If you are unable to download through the link, it is likely because your location is not in China. Please use a VPN or other methods to access.

- Recommended large model: [majic V7](https://www.liblib.art/modelinfo/bced6d7ec1460ac7b923fc5bc95c4540)

- Recommended Vae: [Realistic difConsistency Negative (Pack)](https://www.liblib.art/modelinfo/232ff495f7c14381910dc6f4df78a7ab)

Q: What if there is an error?

A: Just post it in the comments. I will check and respond at 10 AM, 3 PM, and 9 PM (GMT+8) every day.

v1.0.5_20240531

- Optimization: Improved parameters.

- New: Updated interface, now more visually appealing.

v1.0.4_20240529

- Enhancement: Addressed image generation issues to ensure perfect image quality.

- Fix: Corrected the issue with facial realism, which was due to using an anime-style VAE.

v1.0.3_20240528

- Added: Edge blur node for better blending of the model with the surroundings, eliminating any gray edges.

- Added: Discovered that the shadow may not appear when sitting, so I used the double mask. If there is insufficient recognition, you can directly edit the mask in the loading area.

- Fixed: Changed many parameters to make the images appear more natural.

v1.0.2_20240527

- Added: Added a node for high resolution images with a large computational load, to directly reduce the resolution. Default is not connected.

- Fixed: Changed many parameters to make the images appear more natural.

v1.0.1_20240527

- Changed: Replaced openpose full with dw openpose.

- Changed: Replaced canny with anyline.

- Optimized: Improved denoising, feathering, semantic segmentation, hint word inference, and parameters.

- Optimized: More natural smiles.

v1.0.0_20240526

- Added: Initial version.

- Added: Character blending

- Added: Facial redrawing

- Added: Hand redrawing