It was very early when I wrote the image description, and the prompt was of much help. I did my best. lmao I realized couldn’t really figure out what was going on, but I was already halfway through writing the description.

It’s just about. It just needs a little nudge.

You scale the image down and then scale it back up with no interpolation, or use something like unfake.js to snap it to the grid. The author of this model also made a custom comfyUI node to interpret the image directly from the VAE and decode it directly into a pixel image.

Partial occlusions pretty consistently get weird like that.

Whoops. Didn’t mean to leave that in.

The image description is a description I wrote for the visually impaired. The OP didn’t list a prompt or any other generation parameters.

This is not my gen. The OP is linked in the text portion.

It does look weird. Even the texture of the sand is unsettling.

What are your generation parameters? Also, this is more appropriate for !stable_diffusion@lemmy.dbzer0.com. Ask there in the future.

I’m open to feedback.

Yeah, you make really good points, but we get also get the kind of bad faith actor that browses all and downvote community content whenever they see it.

You don’t have to like anything, but consistent downvoting like this without any other kind of participation is indistinguishable from targeted downvote harassment, and isn’t consistent with your claims of blocking.

I’m a moderator on !stable_diffusion_art@lemmy.dbzer0.com, where you’ve never upvoted once, but you have downvoted multiple posts over months instead of blocking the community. This down-ranks submissions, adversely affecting the visibility of the community for subscribers, which is why you were banned.

https://lemmy.dbzer0.com/post/28218497

https://lemmy.dbzer0.com/post/35031819

https://lemmy.dbzer0.com/post/38704142

https://lemmy.dbzer0.com/post/44766239

Check out the model page for an explanation on how it works. It seems to mostly get simpler scenes right. The gallery below the comments has some nice images too.

I noticed that too.

Darn.

Is it a friend?

There are no generation parameters for this image, so there’s no way to tell what the prompt was. The image description is a caption I wrote for the visually impaired.

The model is probably just a bit overfit. More importantly, the image description is not the generation parameters. I wrote myself. There weren’t any generation parameters for this image.

Backgrounds are a problem. They’re hardly ever tagged, but the model can still sort of spit one out without much guidance.