A web service that can actually create images for generating middle divisions is also being released, where `` how to create middle division images of animation '' is discovered with image generation AI



In animation, a cut called '

in -between' is inserted between the original images, which are the key points of movement, to make it look like the characters and objects are actually moving. Twitter user Ao (@ao_love_yuimop) explained that he discovered a way to create such a middle split with `` NovelAI '', an image generation AI that excels in generating two-dimensional images.

I found a method to create animation middle split images including movement with image generation AI, so I made a web service that runs that method (with Yuimoppu video) (everyone can use it) - Qiita
https://qiita.com/ystmg/items/c78c2b3e0bcdf0f77125

chara-zokusei.jp/nakawari
https://chara-zokusei.jp/nakawari

While working with NovelAI, Mr. Ao discovered a way to generate an intermediate image that depicts an intermediate movement from the images before and after it. In fact, Mr. Ao generated 78 middle split images based on 8 NovelAI original images, and the following is a video that was linked and animated. What is drawn is an individual VTuber “ Yuimoppu ”, and it is a video in which Yuimoppu moves around the basketball court with a basketball in hand, and finally decides to dunk.



Mr. Ao actually publishes a service that generates ``images used to generate in-between images'' and explains the creation procedure. First, to generate an in-between image, prepare two images before and after the movement.



Next, specify 4 points in total, 2 points each for the front and back images, such as both ends of the eyes and both ends of the Y-axis coordinates of the body.



Based on this image and feature points, an 'image for generating an intermediate split image' to be input to NovelAI is generated. The generated image is moved so that the midpoint of the feature points input in the previous and subsequent images is in the middle, the scale is adjusted so that it is exactly in the middle, and the original image is arranged so that the background does not become white. , It seems that it is superimposed so that the transparency of the person is 50% each.



By inputting this image into NovelAI and outputting the image with image to image, it is possible to obtain an intermediate split image that is operationally intermediate. It is said that it is possible to create animation of smooth movement by generating further intermediate images as new original images generated in this way. In addition, Mr. Ao explained that in the posted video, he corrected some parts of the image generated by NovelAI that he did not like, and also added a ball.



Related Posts:

in Software,   Web Service,   Video, Posted by log1h_ik