An AI Car Crash Video Clip
Вставка
- Опубліковано 10 вер 2024
- We're in the middle of 2024. There are several websites and apps which provide text to AI video options. I created this clip with the free version of LUMA AI "dream machine":
lumalabs.ai/
I have not tried other portals with this prompt, so I cannot tell who's better or worse.
If you're interested in text to image AI, please see this video I did last year:
• AI Imagines Musical In...
yes , its not good in complex situations yet ! , at the end of the day still we need great cg artist to direct these gen ai, like animators and motion captures
well said!
I've seen some really stellar AI work but after testing, testing, and testing some more it's painfully obvious how they did it: HAMMERING. You have to rerun it over and over and over and cherry pick from a ton of fails. For instance, I had an old picture of a squirrel sitting on a park bench with a nut in its paws. I asked the AI to make it climb a tree and pick balloons from it. Did that twice on Luma Dream Machine. Later on I used the same picture and asked it to simply make the squirrel eat the nut with leaves blowing in the background on both Haiper and Luma Dream Machine. Overall, I ran that same picture through two sites in a total of three different ways. The final shot of it eating the nut came from, guess which? The one where i asked it to make the squirrel climb a tree and pick balloons from it. It was the first attempt in any way shape or form with that picture, and it rejected my wishes and simply made the squirrel sit there eating the nut, quite convincingly. All the rest of the attempts were everything from disappointing to outright nightmare fuel. Moral of this story is you are not getting what you want from these things on the first try, or if you do, you're getting off easy. You have to work and work it, often with just the exact same prompt, to get good results.
thanks for your comment! sounds like an interesting enterprise. my video was the first attempt with a free account. cheers!