saburox
patreon
On the matter of backgrounds
🕑 Added 2024-06-02 19:41:14 +0000 UTCI realize this was going to be an incredibly divisive subject and it's especially long overdue on this side of things, but I figure I might as well address the elephant in the room. I ask that you at least read to the end.
Yes, my pictures have started incorporating some generative AI elements. This has been going on for about a year now. Not as many as some of you might think just from reading that, and not in the ways some of you might think.
I realize I should've written something up explaining this shift way sooner since now it seems like I was trying to hide something. I wasn't; I just wasn't really thinking about it.
So full disclosure, here's what I use it for:
Backgrounds (technically/sometimes)
Not always, mind. Simple things I still do by manually. But when the setting of a piece is more than I'm confident in my ability to pull off, then I've started doing something new. What I'm doing is making an initial background drawing at thumbnail size, and then converting that thumbnail using image to image software, effectively upscaling and enhancing it. Then I refine what comes out, (usually redrawing significant parts of it, or of the thumbnail), and then repeating the process a couple more times before I get a final version that's usable and then I manually do cleanup on that and paint over for things like shadows.
So for example, this:
is the thumbnail that got machine translated to (eventually) become the background for Fledgling Hoard. Incredibly rough, but as you can see the basics were all still there for it to interpret.
It's a cheat yes, but, well...
Before I used to just use visual novel assets I'd bought to accomplish the same thing, which arguably was even lazier. I've always disclosed as such when prompted. (I haven't done this as much lately because I couldn't afford it, there library was limiting, and this ironically felt like a slightly more honest alternative since at least I was providing the initial image data.)
I worry that the amount of layers it'd take to do some of these things, like the above, when combined with the actual subjects, is more than some of the programs I use can feasibly handle. Paint Tool SAI already gets angry if I make the canvas too big. I've had enough drive and hardware failures in the last few years that whenever a file hangs while saving due to being too big and complex I start getting a panic attack and this helps with that, since the thumbnail method means I don't have to do the backgrounds at the full resolution.
The truth is that I've always felt I kinda sucked at backgrounds. Some things are easier than others (I'm pretty confident in my beaches, most of the time) but I think they're a time sink at the best of times. I do try and get better but the returns have always been diminishing and disheartening. I've never been great at drawing large scale props or buildings, to say nothing of how my perspective is often all over the place. It just didn't seem reasonable to spend an extra 30 hours cleaning up a thumbnail for something that's not even the focal point.
So, those are the main reasons I started using tool assist. The irony is that as a result of doing it this way and drawing initial sketches for it to go off of I'm actually doing more in terms of drawing backgrounds and more frequently than I had in years prior. I'm also learning more about how to do from scratch it by using this cheat. With any luck I'll be able to wean myself off of it more and more over time since it's sort of unlocked some things. That's my unrealistic hope anyway.
The other way I use it is
Occasional color picking for the eyedropper tool. This is done by feeding my work (specifically a much scaled down version of the line art) into the machine directly. Then usually I disregard most of what the machine picks out because I already have a color pallet in mind, but sometimes it gives me an idea for something like rim lighting or where to place highlights. I still apply that knowledge by hand with the same good old fashioned tablet I got back in '08.
That's it. The main subject of each pic, I'm still doing that 100% from scratch. Anything they might be holding or interacting with, as well (outside of like, the ground, I guess). Pets, animals familiars, those I do by hand. Background mobs...well those are usually a Clip Studio Paint brush if I'm being honest. Sketch level stuff and line art, obviously, still by hand. And it's going to stay that way. It's really just environments and even those are based on original doodles. Also never for writing. (Not that I write all that often to begin with).
That's why I haven't been checking the "created with AI tools" tickbox over on deviantart. I interpreted that as, perhaps incorrectly, being for when its the entirety/focal point, and since the backgrounds aren't the focal point in my work, and since I was putting in the initial work for the background anyway, I thought it fell more under the category of mixed media and saying it was strictly AI didn't correctly apply. I can see how this can come off as dishonest, and for that I apologize. If the way the wind is blowing says that I need to go back and tag stuff so that it reflects this methodology, I can do that. Again, not trying to hide anything. Whenever someone's said something about a background I've said they're either by someone else (in the cases when I paid for the assets) or in these cases "tool assisted", but I guess I never really outright explained what that meant in a public way until now.
To be clear, I have not, (and will never just) write a prompt into it and posting that generated picture by itself as a finished piece. I do have my limits.
And yes, I hear the response already that this makes me a thief and part of the problem. To that I'd say I've also been stolen from; even if I never dipped a toe into using this technology, I'd still have skin in the game and be a part of the conversation seeing as how I know for a fact there are models (specifically, lora) that are trained on my work.
So I guess somewhere along the line I figured that if I'm feeding the beast no matter what, then I should at least be able to get in line and drink some of its milk too. I guess this does make me me part of the problem though. It's a new world and I'm trying to figure out ways to adapt to it.
The thing is, if you've been around long enough you might remember that way back in the day when I was more hotblooded and and bullheaded (read: stupid), I made some derisive and undeserved and frankly wrong remarks about 3d modelers and ended up burning a lot of bridges that I shouldn't have. In my ignorance of how the tech worked and (because I was angry about the difficulties of coming up with good poses or struggling to figure out camera angles) I'd say things like "they just have to click render and they're done" and then I actually tried to learn how to sculpt 3d models and it was way harder and involved than I gave them credit for. I owe those 3d guys an apology, and I see the same thing happening again and those things I said that were wrong are starting to get echoed. It's not quite the same, and in some cases the argument is more true here than it was there, but in remembering my old attitudes and not wanting to be that person again, I decided to try it for myself and see what the big deal was before making a judgement. Maybe I'm also going too far in the other direction.
Again I know this is probably already damning me in a some of your eyes. I'm not trying to be an AI apologist or anything. I think the problems with how it's being used and rolled out are pretty obvious to anyone with working eyes. But I think I'm trying to land on the idea that this is ultimately a tool and it's here, and that we've gotta deal with it now that it is. There's gotta be some ways for us to use it creatively and ethically and not get swept up or aside by it. So that's what I've been trying to find a way to do.
So, since if you're reading this spiel on this platform that means you've entered into a trust with me to act and produce content to your liking. I'd like to hear your thoughts on what to do going forward.
Comments
Saburo X
In terms of the hair the only thing I didn't use AI at all but I did start trying to use more underlighting to give it depth. Which I guess AI generated stuff does do and I can see how that'd combine with the backgrounds to create that effect. I guess that's another thing I'm going to have to stop doing.
Karvil
I will say that in some of the pieces over the last few months, there started really getting to be enough... something, something about rendering gloss or something, that I started feeling unsure about some them being wholly your work even on the main figures. I think it's partially way AI art tends to make hair so loopy, so work like Trinity Field or Periscopic Proportions started falling into the uncertain valley, especially when I knew the backgrounds were Algo-Gen. Just starts making everything suspicious, even if it's just style changes, eh?
Saburo X
It's not worth the trouble or stress it's causing, so I'm gonna go ahead and ditch it.
Tendra Hinman
While I am intensely opposed to AI art, this is enough of an edge case that it doesn't bother me too badly. You're just using it for backgrounds, and to upscale low resolution backgrounds at that. Wholly generative AI backgrounds might be a bit too far for me, but this isn't that, and I like your work enough that it's hard for me to really hold it against you.
