The everyday blog of Richard Bartle.
RSS feeds: v0.91; v1.0 (RDF); v2.0; Atom.
Previous entry.
9:55am on Saturday, 21st June, 2025:
Anecdote
Midjourney has started to allow the creation of moving images based on static images that it's created. I've had a play with it.
It turns out that it's really good. As with a good many other static-to-moving image AI systems at the moment, people still have a tendency to look as if they're from the Far East in vigorous animations, regardless of whether they do in the original picture; I guess this is to do with where the majority of the training data originated. For normal movement, it's fine. It doesn't have the jelly-fingers in waving that other systems do, although to be fair this seems to be because it makes people wave very slowly.
The guardrails are tighter than they are on static image-generation. Some of the images that Midjourney was happy to generate, it isn't happy to animate. It also likes to think that some images are of children when they're of adults (which it should know because it generated them from a prompt that says they are).
Overall, though, I'm very impressed with it. It's better than everything else I've tried.
That said, it does have some issues.
You noticed the jacket has buttons instead of button holes, yes?
Latest entries.
Archived entries.
About this blog.
Copyright © 2025 Richard Bartle (richard@mud.co.uk).