Grok’s AI video generation is like lazy Sora. All you have to do is upload a picture and it does the rest. You can then refine it through chat. Some of these are the first versions, no input from me. They are mostly pictures from the internet. There are some ones that were still images made in ChatGPT. I think Grok is pretty good at this kind of thing.
A few things to consider:
- The skeletons in the hour glass is an example of a common AI generated video problems. Halfway through the skeletons turn to sand, like in a traditional hour glass. This kind of thing happens a lot, it drops the consistency of the weird thing that happens at first and reverts to what you would normally expect. Related:
- Sometimes there are “errors” in the video. You’ll notice in the pale hag one, the emerald necklace is drawn on at the beginning instead of just being there. Instead of re-promoting to fix that, it’s better to just clip the video in QuickTime of Photos. (a) it’s faster, (b) each time you regenerate, there’s a good chance of reintroducing new errors, (c) each time you reverted, there’s a good chance of it changing the video from what you thought was perfect. Just clip it.
- Sometimes, the video is not so good. But you can see extract good stills from it. For example, the wizard with the exploding crystal ball. The saytr video is pretty good, and you could find all sorts of good stills for an NPC profile picture.
- It generates audio too, easy to ignore if you have your sound off on your phone, as I do most of the time.