Tag: AI
-
Niji・Journey- Inadvertent Asuka
Like her or hate her, she’s so entrenched in the ’90s anime zeitgeist that ‘bots can’t help but give you an Asuka or two even when you’re prompting for something completely different. Not that I’m complaining, mind you, as these came out pretty well 🙂
-
Stable Diffusion 16 Apr 23
This morning, I installed the wildcard manager/dynamic prompt extension in Automatic1111 and set up a test prompt in the following format:1girl, (((solo))), slim figure, beauty adjective, haircolour, upper attire, lower attire, location, time, season, composition The bold items denote the locations of the wildcards in the prompt. As you can see, image generation was pretty…
-
Stable Diffusion #2 12 Apr 23
The pictures below are from two large batches I ran just before leaving home to visit my parents. These are the “niftiest” of the bunch from those most suitable for upload without any inpainting/retouching.
-
Stable Diffusion 12 Apr 23
It turns out that I had performed inpainting and touch-ups on a bunch of pictures and then forgotten about them some time during the beginning of April. To whom it may concern, please enjoy these “rediscovered” cultural artifacts :p
-
Stable Diffusion 10 Apr 23
I let my anime shitmix model go wild. While it will reliably produce an interesting handful, the images resulting from the inclusion of a high number of pose related items in a prompt are too frequently pure nightmare fuel. I think I’ll have to start working with wildcards in order to get the randomness in…
-
Niji・Journey 05 Apr 23
I was testing to see how well NJ could handle hand-accessory interactions (hence the sunglasses), with mixed results; Many of the generated pictures had serious anatomical flaws. In the end, I found that the best sunglasses pics are the ones with hands “near, but not touching.”My next objective was flat, wet hair. NJ didn’t seem…
-
Niji・Journey 03 Apr 23
I messed around with NJ a little more this morning. While it’s certainly better than straight MJ, it still fails a considerably high percentage of the time when generating complex scenes. By fail, I mean that anatomical elements are borked, background characters are completely out of scale, etc.Let’s take a look at some of my…
-
Niji・Journey Testing 02 Apr 23
In my opinion, V5 of Midjourney took a step backwards when it comes to the generation of anime pictures, especially chibi art.Yeah, I could just specify V4 when using MJ, but that takes a bit of the fun away; the latest shiny thing is part of the appeal and I’m a shallow consumer whore. Luckily,…
-
Stable Diffusion 02 Apr 23
I gave a custom-mixed anime model a relatively sparse prompt with a bunch of posing elements (dynamic pose, crazy pose, etc) and let it generate a few hundred images. Some of them came out with a little more dynamism than usual, so I’m calling it a win.
