Stable Diffusion, using the CyberRealistic checkpoint, AnimateDiff, and PromptTravel on Automatic1111. I keep it limited to 16 frames, 8 frames per second. So I only get 2 seconds of low frame rate video, but it is consistent. If I go to 24 frames it does a sudden morph at frame 19. Reading through animation blogs this is a known issue. They say if you use ComfyUI instead of Automatic1111 that you can control it better, but I tried and tried and tried with ComfyUI and wound up with the same, if not worse results, and it took MUCH longer to even generate an attempt. A1111 is much leaner and faster. When using prompt travel enter what you don't want to change on the first line, then start with frame 0: with a description of breast size, then increase by 2 or 4 frames and redescribe. For example, a prompt I used for the above pictures was as follows:
----
25 year old woman, in a living room
0: ((((flat_chest)))) wearing an oxford shirt
14: ((((enormous breasts)))) causing her oxford shirt to rip open
----
The AI will try to render in the woman with smaller breasts, then make them get bigger before frame 15 (N-1 on frame numbers, it starts with frame 0). If you change the prompt it responds better when you change it on even numbers, so you can change it on 0, 2, 4, etc. I have tried that to get smoother growth, going from flat to small, to medium, to large, to huge/enormous. Those seem to be the sizes it accepts and both huge and enormous are similar in size.
I did use some frame interpolation to extend a few videos to get more frames, but it just turns one frame into more and slows the video down like it is slow motion. The one you attached looks like the artist did that. I would love to get it to 24 frames per second and get up to 4 seconds of action, but too many morphs and it also makes the videos too large to post as GIF files and I don't want to start posting elsewhere to accommodate video files. There is something else that installs with AnimateDiff that works the weird morphs into smoother transitions I used, but it can still be a stark contrast in background or composure and just looks weird.
I again am thinking of hanging up my hat as far as AI art goes. Because I am a perfectionist I tend to go down rabbit holes trying to make the videos better and better and it eats up way too much of my time. I am looking to change careers and also working on a board game. I need my time to concentrate on those things instead of making spank material on the interwebs.
Punchy