Hey Viktor, looks cool! It looks like it uses some image-to-video generation model. I would be interested to know which one you used, how you selected it, and whether you did some training/fine-tuning yourself in the process.