Version 2 says to use this prompt:

lyy666，快速切换， [DESCRIBE THE NEW SCENE]

For example, in their video, the description of the new scene started with this:

"lyy666，快速切换， At 3 a.m. in the back alleys of Kowloon, Hong Kong, the heavy rain has just stopped. The water-logged cobblestone streets reflect the shattered neon signs—“Red Wine” is a crimson pulsating spot of light, “Tea Restaurant” in a blue mist. A woman in her late twenties leans against a rusted telephone booth, dressed in a dark green silk cheongsam. The rain-soaked fabric reveals her collarbone and waistline. Retro cat-eye glasses slide down her nose, and her lips are a deep red like aged lipstick."

and went on in even more detail 3 to 4 times this length.

Jellai Wich video?

cldoerick131 This one: https://civitai.com/images/94155040

Did you have success using it? What it does exactly?

From what I understand, it makes the base image switch faster to what we prompt?

cldoerick131 Looks like It helps you change the scene and still keep consistency with the characters/objects used in the first frame

By having a scene change within one generation it is more likely that the characters will look the same. And this makes the scene transition faster. Meaning less of the limited frames are lost to the crossfade.

就是用在i2v流程中，实现用你的参考图快速过渡（看起来就5-6帧）到你想要生成的视频，而这个视频可以和你的图片不是连贯的，比如你有张穿T恤的美女，但你想生成她穿古装，就可以用这个，提示词直接说“古代美女在干嘛”，生成视频可以保持脸部的一致性。只浪费前5-6帧， 但可以省去类似stand-in那种复杂的流程。