oligetcetera-deactivated2023072:
I think a decade or two ago it used to be common for artists working in Photoshop to explain that “no, this is really ‘painting’ just as much as traditional media, it relies on the same skills”. But now the AI art tech kindof vindicates the traditionalist-purist side?
Like, it seems clear that generating images from text prompts isn’t painting, but at the same time there’s a continuum, and out of the many labor-saving devices like perspective tools, custom foliage brushes, posable 3d models to paint over, etc, it’s hard to say where the line is. So the new tools retroactively make the previous ones suspect…
maybe i’m just basic but this actually seems pretty simple? all of these are just different media. goache is not watercolor is not photoshop. some skills transfer more than others. (appropriately enough the paintbrush function on photoshop transfers a lot with trad media paintings but a lot of the more specific and interesting new things you can do with it are less so, obviousy.)
some media have a much lower skill floor for producing something passable - despite a lot of effort i can’t really reproduce someone’s likeness reliably with pen, but i absolutely can with photography despite never attempting anything like that. really good photography is a skill, same with promptcraft. (hic rhodus hic salta, if you think anybody could make those balenciaga edits, same as jackson pollack or whatever)
that said, i feel like even with the fancy tricks pre-AI digital illustration has a skill floor more similar to traditional paintings’ media than to photography or generative ai, so if people want to draw a convenient line there for identity purposes that seems fine to me
I feel that just saying “if people want to draw a convenient line” is treating the matter too lightly! People respond differently to paintings and photos. People on social media respond really differently to paintings and AI works. :) Unlike gouache versus watercolor there are value judgements involved. So what exactly is valuable?
🎨 Back in the day, when Leonardo da Vinci argued that painting is more noble than sculpture, he said one of the reasons is that the painter uses linear perspective.
Now to return to the proposition concerning what was said about low relief, I will say that it involves less physical fatigue than full relief, yet it is a much greater investigation, for the proportions interposed by distance between the parts of bodies, from the first part to the second, and from the second to the third in succession must be considered in [low relief]. If you have not considered these [distances], perspectivist, you will discover no work at all in low relief that is not full of errors in cases when greater and lesser relief is required for the parts of bodies as they are closer or farther from the eye. In full relief there will never be any such error because nature helps the sculptor, and for this reason whoever does full relief is deprived of such difficulty.
If you use 3d reference figures, you are similarly deprived of the difficulty. One specific skill is automated away.
🎨 A traditional painter spends a bunch of time learning to mix paints and will say things like “one of my favorite pigments is alizarin crimson”. In digital this is obviously irrelevant, and instead people will e.g. apply color to entire areas by messing around with the “curves” or hue/saturation tools. Actually, digital photographers use the same tools to postprocess their pictures. This makes the process a bit less premeditated and more experimental: instead of mixing a particular hue you can move around the sliders until you happen to come across a good one.
Or, you can use AI tech to go directly from line art to colors. I guess you can try this multiple times until you happen to come across a good one. (Someone on the youtube page comments “OMG~ i was struggling with coloring and here we are WOOOOW~ Thank you~”.)
🎨 One case I had in mind. Some acquaintance was looking at a scene from a Makoto Shinkai movie and exclaimed “so beautiful, it looks like a painting!” and I replied that it’s literally a painting. But is it? Back in the 1980s it would have been created using physical paints, but now it’s drawn digitally. It makes heavy use of photo references. At least in some cases I would guess it’s drawn/traced directly on top of photos. In any case you can make similar art that way: by googling I found somebody’s “tutorial on painting over a photo to turn it into a Makoto Shinkai style anime background”, or a another similar tutorial.
The process described in the tutorials doesn’t rely on drawing skills or perspective construction, because you trace. It does require a sense of color, but in a curve-tweaking way that is more like modern photography than traditional painting. The rendering is quite mechanical (“now we draw lines everywhere”). I think we have traveled quite a bit towards the “not painting” side of the spectrum even before we bring out the modern AI tech (“AnimeGAN: The Photo AI That Can Turn Any Photo Into a Shinkai or Ghibli Anime”).
So how many tasks are automatable before the nobility of painting is in danger? I guess my point is that nobody really had to think about this until now, because digital and traditional painting were obviously similar enough to count as “the same thing”. But going forward I think the way commercial images are produced will become radically different: some mix of rough sketches, posed models, text-based prompts, extrapolation from mood-boards, etc. And I suspect people will want to distinguish “that thing” from “painting”. Maybe we have already overshot, so some of the practices that we accept now will be considered “not painting” in the future.
Lots of very good thoughts here, and I think as productive realities slowly wipe away “ai art: yes or no” debates I think instead these sort of “what do we actually value when we say we want the human element in art” questions are going to be the more fruitful ones, which this is definitely pointing at.