Hacker News new | ask | show | jobs
by maalber 540 days ago
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.
1 comments

Hey! Yes, it's using image-to-video. I went with a model by KlingAI as it seemed to produce the best results from my tests.
I found their API quite expensive, I wonder how you go about it.
3 for $25 is how