A workflow for using LLM to make short videos. You just need to give a concept and do nothing. Chinese or other language can be used for prompt.

The workflow is to make the AI think it is a director to make a film, which requires eight scenes, generates corresponding pictures, and lets the AI add special effects to each scene. Finally, the special effects instructions are added to the PromptSchedule to generate a complete video.

The current revised workflow version no longer takes up a lot of VRAM, and the local default uses the smaller phi LLM model and Dreamshpaer SD 1.5 model.

==

一个使用AI制作小短片的工作流，你只要给个概念就可以，可使用中文。

工作流是让AI认为它是一个导演，需要6个场景，产生相对应的图片，并让AI在每个场景里面添加特效，最后做成特效的指令加入PromptSchedule，以生成整段完整的影片。

目前修正的工作流版本已不会占用大量VRAM，本地预设使用较小的phi LLM模型及Dreamshpaer SD 1.5模型

==

Example Concept：

1、Deep in the universe full of flowers and firework

2、宇宙中開滿了花朵

3、girls wearing kimono in summer

4、Nightstreet

5、The birth of earth