DiffusionBee AI image software

@MacSparky posted about DiffusionBee:

I played around a bit with variations of the prompt he was trying.

Order matters. Putting “macintosh computer” at the beginning of the prompt made it more prominent in some of the images than leading with “bonsai tree.”

“macintosh ​computer with a bonsai tree on top”

Adding “desktop” to the prompt produced striking results for the computer part of the image, at the cost of less (or no) bonsai trees.

“macintosh desktop computer with a bonsai tree on top”

More broadly, DiffusionBee is not at the level of “type what you want and get an image of it” right now. It’s more “type what you want, generate a lot of images, try a bunch of variations on the prompt, and maybe a few of them will be what you want.”

1 Like

There’s a bunch of web based tools now. Dall-E, was one of the first, but a bunch others are available.

Even Microsoft is adding one to their new “Microsoft Designer” product.

It’s the new “shiny thing” every company thinks they have to have.

Even Facebook, shudder, has a version in their labs that generates video (just a sequence of photos, after all).

No need to install anything, and IMHO, the web-based tools are improving much faster as they keep adding to their data and algorithms.

Also, if you don’t have a M1 Mac, the web based versions are really the only option.

2 Likes

Shudder, indeed. “Mark Zuckerberg showing concern for users” [ERROR! BAD INPUT!]

I think I will start applying the same rules to tech as I do for TV/movies — the bigger the hype, the less attention I will pay to it.

Along similar tech-babble lines, I have warned my younger friends who have never been through a big economic downturn, if your job title is something your Grandma doesn’t and cannot understand, better start polishing up your resume now.