The purpose of this workflow is to solve the following issues:

Character consistency under different actions

This workflow requires approximately 4 action images(like those in assets), which can be changed according to your needs.

Explanation of each group:

1. Prompt: Enter prompts and adjust them automatically through LLM

2. StartImage: Initial image to establish the initial style, generating two styles by default. The basic actions of the styles are selected from the four action images, and you can control which image to use

3. PreviewChooser: Choose the desired style and click "Proceed with..." to continue the following tasks

4. BatchPose: Action image area, the default switch is 2. Directly place the action images here. If it is not a skeleton image but capturing a person's actions, please set the switch to 1

5. BatchAlign: Generate images of the character in different actions

6. FaceDetailer: Face restoration

If character consistency can be maintained, the character's actions can be placed in Toon to connect them into an action video (like the sample short video), or use other video platforms to create a series of videos.

In principle, the face in this workflow will be consistent. If you want to change the face, you can add nodes (like reactor) yourself.

Compared to workflows that directly generate finished products, problem-solving workflows are less common. The workflows I have recently created do not generate finished products but solve the problems I encountered.

==

这个工作流的目的是解决以下问题：

不同动作下的角色一致性

本工作流需要动作图片约4张，可视自己需要调整。

各群组说明：

1、 Prompt：输入提示词并透过LLM自动调整

2、 StartImage：起始图片，用以建立初始造型，默认产生两张造型，造型的基础动作由四张动作图片中选取，可控制要使用第几张图片

3、 PreviewChooser：选择需要的造型，按Proceed with…接续以下的工作

4、 BatchPose：动作图片区，默认开关是2，直接放入动作图片，若并非骨架图而是要捕捉人物的动作，请把开关调成1

5、 BatchAlign：产生人物的不同动作图片

6、 FaceDetailer：脸部修复

若能保持人物足够的一致性，可将人物动作放入Toon使其串接为动作影片(如范例短影片)，或使用其他影片平台，以做出系列影片。

原则上此工作流的脸部会是一致的，若要换脸可自行添加。

相较于直接产生成品的工作流，解决问题的工作流比较少见，最近制作的工作流并非产生成品，而是解决我遇到的问题。