This was a learning experience for me.

Started with this:

And ended up here:

I used NovelAI’s imagegen.

I used 6 different prompts aimed to get the starting lady further and further into a transformation sequence. Ended up with 104 images. I then used RIFE to interpolate the images and give me my frames. Finally I used ffmpeg to convert the frames into a webm.

Issues I noticed:

• The sundress briefly converts to a long sleeve dress towards the end. I should have been more careful while inpainting
• The frames seem to “stick” to each primary image too long, I’m not sure how to make it smoother yet. Maybe a different tool?
• A little bit of artifacting here and there
• The imagegen did not want to do red_skin, like at all