Testing Kling AI's image-to-video model on generating a third jump shows the background improves but body mechanics fail. The legs won't cooperate, looking like a second jump or backward motion instead. Multiple prompt tweaks and a ChatGPT generator don't fix it.
So you can see the dramatic improvement here in CLING version 1.6. Unfortunately, the body mechanics of the third jump are just not adhering to the prompt well here. So you can see that the background is still looking fantastic, but the mechanics in this one are showing a running motion and resembles more of a second jump and not a third jump. So here I tweaked the prompt a bit and finally I'm starting to get the body to move more in the motion it would in a third jump, but it's not quite there. Here you can see the upper body looks much more accurate with the arms extending, but the bottom half of the body now is totally out of whack and not accurate. Look at the muscle movement on the legs. It's pretty impressive. However, I changed this prompt and it didn't get any better. It only got worse. So it was time to try something else. So now I used an enhanced prompt generator with ChatGPT, but you can see it just really didn't improve anything here. So I'm still using the enhanced ChatGPT prompt generator, uh, and I'm still telling it to do the third jump, but this resembles more of a second jump and it's just not really working. So I finally tried one last time. This does sort of resemble a third jump, but I'm going backwards and it's just not quite working. So I pretty much gave up at this point.
Ultrawide Monitor with PBP (Picture by Picture) demo using Logitech Flow with PC and Mac
