I thought it would be interesting to document how I am currently creating with AI, and how I am achieving things like adding X's into my artwork. Essentially, this is an article about editing AI artwork, by adding to and subtracting from your images...
Before I get started, I want to emphasize that this article is about how I am currently creating with AI. It is NOT an article about how I think YOU should be creating with AI! The creation process is as individual as the creators are, and each of us do things in our own special way. There is no wrong or right way to create!
That being said, I know there are at least a few people who are curious about what I'm currently doing, so here is an in depth look at my current creation process, from start to finish. As you'll see, I do a lot more than just click on a 'generate' button! There's a lot of editing going on, by adding to and subtracting from my AI artwork.
Today I will show you how I created this piece:
As is often the case with me now, I had no idea WHAT I was going to create when I started. When I made AI images from pure prompts, I always knew beforehand what I was doing, but now my creation process tends to start with a visual prompt, so I went to Google Images and started surfing there.
I can't remember the specific search I did at Google, but I think it was "portrait of a person on dark background". After a while I found this image to start with:
Obviously, it is AI generated, as we can tell from the hand with only 3 fingers! Sadly, I have no idea who created it, or even from what website it is published on. But it already had the really BIG HAIR that I'm in to, and I liked her pose and wanted to use it.
I was actually hesitant to use it though, as I knew in advance that those hands would likely cause me a lot of headaches in the creation process, but I went ahead with it anyway.
The first thing I did with the image was resize it to the dimensions that I wanted my finished work to be. Also, since I currently have this strange X fetish going on, I added a big X onto her dress!
Then it occurred to me that I'd like to add some highlights to her hair, so I drew those in as well. I did the resizing, the added X, and the highlight drawing with IRFANVIEW, a free image editing program. Of course any half decent editing program will allow for the same things.
OK, so I now had my first visual prompt to start playing with in AI.
I took it to Playground.com to get started. The reason I went there is because I can generate hundreds of images daily with my account, and doing so on some other AI sites could cost me a small fortune! Having practically unlimited generations in AI is an artist's dream come true, as you can keep redoing things until the AI does what you want. You can read more about that HERE.
As I eXpected, my first few renderings produced less than stellar outcomes. I created a written prompt to eXplain what I had in mind. I then kept changing my prompt with each new generation, to coincide with what I wanted as well as to what the AI was creating on its own that I liked. Here are a few eXamples...
The first image shows my first rendering. The second shows my fourth rendering. The third shows my tenth rendering.
I asked the AI for a 'shiny leather dress' which began turning silver after several renderings. As you can also see, the AI didn't really understand my directions for highlights in her hair, and instead started giving me silver 'things' located there.
So I went with the things I liked and changed my written prompt accordingly. I asked the AI to give me variations on the "X" and to make her dress silver. I asked for 'silver ribbons' in her hair and also for a cloudy sky background.
I did several more renderings with miXed results...
When I came to the above image I knew it was the one I wanted to continue with. I just LOVED the way the AI incorporated the letter X into her dress design! But it was pretty sloppy too, and had a LOT of problems.
As eXpected, her hands were horrible. They each had 5 fingers and one of them had 6 fingernails! We all know I love X's, but there were way too many here even for me. That silly one on her head had to go.
So I took the image to LUNAPIC.com, which is a free online editing website. Several months ago now, some nice person at DDG told me about the site and I have been using it ever since. I wish I remembered who it was, but of course that private message there is now long gone.
Anyway, LUNAPIC has a wonderful tool called 'Smart Object Removal' which you can find under the EDIT tab there. I suspect it is also another AI application, as it does such an amazing job at what I use it for.
LUNAPIC has 'saved' dozens of my AI images over these past few months, as I can erase practically anything there. I remove tiny imperfections, as well as large objects like arms and legs when necessary. If you haven't tried it, then do so. It's amazing and it's FREE!
So I took my image there and removed the big X on her head, as well as the troublesome eXtra fingers, then ran the edited image back through the AI again as my source image.
Before LUNAPIC
After LUNAPIC
So now I felt like I was really getting somewhere. I liked the silver bow in her hair and the silver X dress, and I had the right amount of fingers, even though the hands were still very crude in appearance.
Then I realized I had accidentally left one of her fingernails from the finger I removed, and I didn't like that little X on her dress under the big one, so once again, back to LUNAPIC!
Ah, yes! Much better, indeed! Now I downloaded the image and manually sharpened it in IRFANVIEW. Then I'd take the sharpened image back into the AI and use it as my new visual prompt and render it again. I likely did this a few times, as it really helps the AI produce more detailed and crisper images.
Now it was time to start working on her hair. I had that big black area in the center of it where the X was, and her hairdo was lopsided! I didn't care for the one side being bigger and 'poofier'.
Back into IRFANVIEW I went for some manual additions...
The image I took into Irfanview this time.
I copied an area on the top left part of her hair and ribbon, then pasted a mirrored image of it on the right side.
Then I used a fine pencil and various colours to draw in the hairdo more to my liking, and to show the AI what I wanted it to do.
Now I took the 3rd image above back into Playground.com for more renderings to see what I might come up with. After a few tries I came up with this image...
Now I had no big black void in the center of her hair, a more balanced bow, and a bigger balanced hairdo with some colour in it!
To be honest, I was quite happy with this image and thought I was done with it. But the hands were still bothersome to me. So instead of continuing with more rendering at Playground, which I thought I had reached my limits with, I thought I would instead take the above image over to the Deep Dream Generator to see if I'd have better luck there with the hands.
I find myself doing this a lot these days. Taking an image made at one AI site and then uploading it as a source image at another. It takes a bit of practice, but if done right you can end up with an image that has the best qualities from BOTH sites/apps, and one that doesn't look like it was fully created at either! You can come up with some interesting and unique 'blends'.
So off to DDG I went. I ran the image through the DigitalDaVinci filter there and came up with this...
I wasn't overly impressed. DDG completely changed the expression on her face, and I preferred the seXier and more mysterious features at Playground. Also, the hands were STILL messed up, though they were an improvement from many previous versions.
Then I noticed that DDG had a new feature in the AI Upscaler section called "Face/Skin", so I took my new DDG rendering in there to try it out. I used the Face/Skin mode with the Realism engine and after a few tries, it pooped out this image for me...
FINALLY, I knew that I was done with this image, and frankly was very pleased to be so! I lost the seXy face I liked, but this new face was still pleasing to the eyes. I had my fancy BIG hairdo that was balanced and had some colour to it. I also had a semi-balanced BIG silver bow in her hair, and a singular BIG letter "X" that was a striking part of her shiny silver dress!
But, maybe best of all, (to me), was she finally had half decent looking, (somewhat believable), hands. They had the correct amount of fingers and even fingernails on them!
Yes, there are still some things that I see 'wrong' with the above image. Those earrings are horrible and need to be smooth and shiny chrome hoops, (or to be just removed altogether by my good friend, LUNAPIC).
There's that spot on the right side of her bow, between the top and second rounded ribbon, I could easily fiX that if I tried.
And my clouds somehow became sort of 'digitalized' with all my renderings, and a shiny spot appeared where the sun peaks through them in a most inappropriate area. I'm sure some sweet talking in my written prompt would coaX an AI to fiX that up for me!
BUT NOT TODAY!!!
Sometimes you just have to know when to stop and say, "ENOUGH!", and remember that with AI you can always return to the image later and keep working on it, which I'm sure I'll do in this case. But by now I was totally sick of looking at this girl's face, (no matter WHICH face she was currently wearing)!
I worked on this image for 4 days. I know I worked at least an hour on it for each of those days. I used 2 different AI sites, and made 99 separate renderings of this woman on them.
I made 92 renderings at Playground.com, (because I could afford to), then made an additional 7 renderings at DeepDreamGenerator.com, (including using the AI Upscaler's Face/Skin mode). I went to LUNAPIC.com countless times to remove the things I didn't like from my piX, then used IRFANVIEW another countless times to add the things I wanted to them. My picture took me many DAYS and many HOURS to complete!
Now, as I mentioned at the top of this X-File, I don't think YOU should necessarily create this way. It's just the way I happen to be creating with AI at the moment. Really, this is a document mostly for myself to revisit in the future, so I can remember what I was doing at this moment when I look back at it years from now. The things I'm doing with AI currently are sure to continue changing as time goes by.
Also, for those few of you who were wondering, now you know the torture I put myself through, ha ha! But the only reason I am creating this way is because I WANT TO. It is creatively pleasing to me, though I do tend to get equally irritated as well during the process!
All this 'work' I'm currently doing to make pictures with AI does seem to be totally at odds with the whole concept of quick and instantly gratifying images provided to us with the click of a button. For me personally though, it is giving me the satisfaction of putting work and thought into my final image, and that's very important to me.
I think what's best of all though is what I'm creating. I'm making interesting images that are a bizarre blend of my own imagination that has been equally miXed with the AI's imagination.
The 2 of us, (or 3 or 4 of us), seem to work quite well together!
I hope you enjoyed this little peek into my satisfying yet sometimes irritating creation process.
Best Wishes,
"X"
June 28th, 2024