🔺 Devil's Toy
A video created using Hailuo AI 😲
It's astonishing how good it is... Obviously, it's just for fun and not meant to be "fine art", but it's so fulfilling to see your creations come alive that I felt like sharing it :D
I've been experimenting with it all day and there are definitely some best practices to follow as the AI tends to make weird choices without proper guidance. I found it's better to turn off the auto-refinement of the prompt, then describe what the image actually depicts, and finally how the animation should unfold, e.g. "A young man is doing yoga while leaning against a wall, he's holding his position while turning his head to smirk at the camera".
I'm on the free plan for now and there's a long wait for each video, so it's very hard to experiment with things properly. I tried some silly stuff like "raising his butt to touch his head" but several attempts didn't quite do the trick.
It also struggles with faces, so it didn't capture Roberto's face accurately. In fact, here's the videos I got right from Hailuo AI, and you can see how it turned out. 😂
I had to go frame-by-frame, fixing his face to look like him, but I must say I'm too happy with the results to be that spoiled fella who whines about their internet not working on the plane while flying 9km above the ground :P It's still in the beta so I'm sure it will be even betta over time XD
And I noticed it because I have one maybe obvious idea but one you might not use. When you create such video you could go frame by frame and select one for wholy new image than it was one you started with. Then you could either use it as it is for new image OR feed it to ai and create another video and in a such way you could "chain" videos to create longer ones with more profound changes!
Just and idea in any way thanks for sharing. Your skills show as without fixed face poor Roberto looks bit like he is already experiencing some satanic stuff X-)
Shifty
If you begin from the last frame of a previous video to create a new one, the AI wouldn't understand what motion was originally in the prior video, so you can't really stitch them together smoothly.
The idea to pick frames from such AI-generated videos is a good one! So far I haven't seen any interesting frames worthy of turning into a static artwork, but I'll keep this idea in mind :D
Also, tried to add the video controls on the site, so hopefully it's better now :D
Ah I see it is unfortunate ai cant chain like that.
Controls work good in new browsers but in that old one click opens full video page :)
Shifty
Clicking on the image opens the full view but I guess it's a bad idea to do the same for videos XD I'll fix it!
____
Divine! Thank you!!!
____
Thank you to you too! 🙏🤗
GREAT work yuni👏