Merry Christmas to all, and to all a good night.
Comment
Baby It’s Cold Outside, and while Alex does love a good Christmas hike through the Minnesota lake country, she does want to come inside to warm up.
I enjoy showing people the things I create. Either stories or pictures. But there is a certain number that overlords people. The more pictures I post, the less people look at them. I’ve tried around a dozen, five, four, two or and I’ve tried just one. Ironically, as the saying goes, less is more. More people view and interact with the pictures I publish when I post two images rather than five or ten. Four seems to work well, but two is still better than it. So here are two images of Alex getting ready for Christmas.
Alex from jackofharts.com generated by JuggernautXL 5 AI at dezgo.com
Baby It’s Cold Outside, as an old song says, and here is Julie, looking for an opportunity to warm up.
There are around a hundred pictures in the Christmas render archive I took these two out of. I probably generated a thousand renders to get them, only keeping the ones I really liked. This actually doesn’t take a lot of my time since I run the renders between watching shows, reading books, or writing my own stuff. I check in every hour or so, when getting up to do things, or going to bed, or waking up, look at the images, see what I like, save them, and run another set. The hardest part is paring them down to the two to publish.
Julie from Jack of Harts.com, generated by JuggernautXL 5 AI at dezgo.com
Step Two of getting good character reference art is easy peasy. No brain needed. Just start a generation and walk away. Come back, save what you like, and start up another run. Usually 1 or 2 of each batch are just bad art because the AI went spastic. But the better AI generators can take a final prompt and give you 2 or 3 out of 4 that look good. Once I saved 50, 100, or more images, I would do a more in depth pass, deleting the ones that were good but didn’t match my mind’s eye. Usually 25% failed that pass. Sometimes 50% if the prompt resulted in particularly erratic results. Once I had between 50 and 100 images I liked, a third pass would find the 1 closeup and 1 torso shot I wanted for the final reference art. That all took about a week.
Getting good reference art from the AI generators using text prompts is like coding computers or writing a story. You start with the setting and the character, telling a visual story of what you want. The AI spits a picture back. You change the words and try again. Move the words up or down in priority. Maybe even add a + or – to the word to really reinforce what you want or don’t want. I had 3 or 4 PCs with 4 windows open, running generations that I would check every half hour or so to see what they gave me. Between doing other stuff around home, that usually ends up being 20 to 50 images per PC a day to analyze. It usually takes me 2 or 3 days at that rate to find the prompt that will reliably give me the character and setting I want. That’s Step One.
Forge of War on Amazon
Angel Flight on Amazon
Angel Strike on Amazon
Angel War on Amazon
Wolfenheim Rising on Amazon
Wolfenheim Emergent on Amazon