Purposes:
Whether you're intrigued by Image Diffusion (more commonly known as "AI"), against it, or non-bias as my blatant-self, and in case someone's curious.
For the intrigued, it's fun watching how this thing works each time. Or as a demo of a process. Nothing too specific, just the feels of it.
For the anti-s, it's to show that it's not just putting some texts and letting it do everything else, there's a fair amount of human work involved (at least in this particular process), and it's not much more than a brush set and a Photoshop tool for me. So that's about 1 out of like 7 reasons I can think of to hate this mitigated, and hopefully you can enjoy the thing a bit more.
For the unbiased, well, you're here for the same reason I'm here: why the heck not. I'm in a rambly mood. Maybe you're in a reading mood. I dunno.
Apps:
The app used is Krita, a drawing app. Found out it has a plug-in you can download and install that does the diffusion, and then compatible models that work for its specific software to download so it can reference, while letting me have similar composition control over it as if I'm drawing or photo editing.
Requires a bit of technical reading ability or lots of button-mashing.
I'm the latter.
Image Explanation:
The GIF is made with leftover layers after working with it.
The average chance of it giving me something I can work with is 1 out of about 20.
There were 265 layers.
This does not include all the undo-s.
The gloomy atmosphere and rugged texture work. It would take hoooouuurrrrs just to do a face for me. I'll be bored to death if not the bad posture giving me a seizure. And I'm not properly trained to do most of those textures, which then I'll have to add that training time to get these atmospheres.
It saves time. A smarter way to work when not having enough time.
The nasty room it started with, and good fog outside. Fantastic work.
Also, as mentioned before, fun to watch examples. Occasionally inspirational, even, when it does something that I wasn't expecting, and it looked decent enough for me to work on it.
Then the things it absolutely cannot do (on its own):
Hair on the face. I drew that.
Completely drenched hair. Even with a specific model installed for it, it won't. I gave up on this one.
Longer hair with ponytail. It did the ponytail beautifully. Won't make it longer. Drew that.
Tentacles with mouths or eyes on their surfaces. I gave up.
Barely got a tentacle with a head and a mouth on it. But I want that head facing away and on the crotch. It won't do that. There is not enough reference on the entire internet for these things to know how to do that, I think. So I had to draw it, and keep stopping it from changing into a weird penetration.
I wanted more mouths and some freaky deformed human features all over the background flesh walls. But I had to compromise. At least it layered all around it after I painted red on everything. Lost the darker atmosphere after that, though.
Arousal face as a whole. It does that well if it's just a portrait of a woman. For some reason, the face breaks when there are tentacles around. Had to change mouth, eyes, and brows separately (it did glasses beautifully, though).
Wanted more post-apocalyptic gears around her. But, nah, that's about as good as it's going to get. It just keeps changing my drawings into weird strings or melting tentacles.
Wanted the pants to bulge a bit more for the tentacle head that's in it. It won't do that. I'm quite happy with the rips, though, and got lucky with the zipper.
The hand. My gods, I got lucky with that one. I could not control it.
Sometimes it refuses to fill the cropped area with what I want, like the floor situation. So I have to just tell it to basically reset everything in that area, and hope for the best.
Conclusion:
It's a toy.
It's fun.
But now my expectation went up, and I've scrapped the last 3 images that I attempted after this one because it just won't do what I want it to do. I can't say I want to do that too often.
The way it works reminded me a lot of how I do Blender animation, video editing, drawing reference research, or grammar-checking. So now I miss those and I want to go do those.
Or just draw. I've been thinking of simpler line drawing. You know, dropping gorgeous textures and atmosphere, instead more storytelling and control.
I might flick at it once in a while and see if it gives me anything interesting that makes me gotta-work-it.
Otherwise, one of many tech-freaky productivity apps on my shelf.
Also, the GIF on top starts from the end-to-beginning for storytelling purpose with a recognizable reference.
Thought it might be easier to study with a beginning-to-finish version, so I just had it made while I'm at it:
ps. Originally, I named the post "GIF Deconstruct (and Reconstruct)" of the thing, since GIF was the main point.
Then I wrote the long thing. So... changed.
Comments
Post a Comment