Model:RealVisXL V4.0

==

A Workflow for Automatically Applying Masks and Composing Characters

Based on SDXL, using WD14 for prompting, Yolo for masking, and LineArt for fixed poses and styles.

This process is divided into three main groups:

1.Mask: Use Yolo + Mask to draw masks for three characters.

2.Controlnet: Use LineArt to control the image. Note that LineArt must be set between 0.75 and 1.

3.Adapter: Use IPAdapter Params to combine the three character portraits with the background.

Required Images:

1.A reference pose image for the group photo (the top gray box).

2.Three individual character images and one background image (the background is the blue box, and the other three colored boxes are for the characters).

Due to the use of LineArt, it is recommended that the character and pose images do not differ significantly, and it is best to use half-body shots.

When starting, it is recommended to only activate the Mask group first. After confirming the masks for the three characters, proceed with the subsequent operations.

Notes:

This workflow can achieve a realistic effect, but the face shapes might not be very accurate. It is necessary to use PULID to change the face shapes for a complete result. The clothing styles can only be roughly captured.

This method can also be applied to four or five characters. Simply copy the nodes with the labeled colors and connect them to the Params to logically extend the process.

==

一个自动绘制蒙版套用人物的工作流。

以SDXL为基础、WD14提词，Yolo蒙版，LineArt固定姿势及样式。

基本分为三个群组

1、 Mask：使用Yolo+Mask绘制三人蒙版

2、 Controlnet：使用LineArt控制画面，注意此处LineArt必须于0.75-1

3、 Adapter：使用IPAdapter Params组合三人头像与背景

需准备图片共五张：

1、 合照参考姿势图片(上方灰色格)

2、 三个人物与背景图片(背景为蓝色格，其他三个颜色为人物)

由于使用LineArt，人物与姿势图片不建议差距太大，且以半身照为佳。

开始操作时建议仅开启Mask群组，确认三人的蒙版后再进行后续操作。

==

此工作流真实感可以，但脸型会不太像，后续需套用PULID融合脸型才完整，衣服则只能大概抓形式。

使用此方式若有四人、五人均可套用，将有标注颜色的节点复制，连接上Params就能依逻辑往下推演。