V2

Tried burning the image into the lora so it can try to generate the pose finger at different locations - fail

Instead, I used img2img on the four different frames of the video with canny control net to force perfect fingers.

<- get img2img clips yourself using ffmpeg using chatgpt asking it how to extract video to png for only 2 seconds

1. img2img (^ total 4 frames)

2. control net

3. video editing software any

4. flow frames

should be four images but to loop you need gif

1,2,3,4,3,2

V1

It is not an animation, it is a pose concept

images size

• 832x1216 or reverse

• or 1024x1024

doodledance, index finger raised

• outstretched arms (optional can cause errors)

描述:

训练词语: doodledance, index finger raised,