a-flyleaf.github.io

untitled post

September 12, 2022

It took me a solid ~5 minutes of running Jekyll, closing it, clearing the cache, restarting, rinse and repeat, before I realized this post wasn’t showing up because I was refreshing my live blog page and not the localhost.

Anyway, assorted musings with minimal relation.

untitled subheading

Screenshot of an unread Gmail message from Zell Liew, titled “Designing a focus style” and dated to October 2nd, 2019.
Alas, I still use Gmail.

Ages ago I subscribed to a developer’s email newsletter; checking my message history reminds me that I was probably aiming to take his JavaScript course. (A majority of these emails are unread.) For whatever reason, the one pictured above, “Designing a focus style,” was singled out among Gmail’s automatic tabs—and there it languished, unopened, since being sent in 2019. Because I thought it was about making your own time management technique, and figured I’d read it later.

I opened it recently just for the hell of it, even though I had no intention of sitting down and brainstorming a new time management technique.

It is not about time mangaement.

I feel like this is one of those things that characterizes me as a person.

blog-blogging

It’s dawned on me recently that my deviantArt descriptions are the closest thing I have right now to a blog, in terms of updating the public on my life and pursuits and such. On one hand this is fine; I’m content shouting to a void, don’t want all my personal beeswax* out there for the world to see, privacy etc. etc.

On the other hand, it’s deviantArt, and there were a good few hours a while back where all descriptions suddenly appeared to disappear. With all the promotions and monetization ads being shoved in my face throughout the site, I should really not rely on it for a reliable archive of my art and associated description-writing.

I rely on it anyway because #hashtagConvenience. Damn the bugger. I’m getting better at sitebuilding, so maybe someday, but not today.

Today, though, this exists.

oh yeah, the main thing I wanted to get at by bringing this up

Last month I didn’t draw much so my roundup was one of the shortest since I started doing roundups. But I stil did a lot of digital work, just in Blender and/or site-related! It’s just not the kind of thing that lends itself to what are essentially sketchdump-collages.

Relatedly, it irks me somewhat that, while I’ve made an effort to caption images throughout ↑ that site (and 404 before it, albeit to a lesser and much less convenient* extent), I haven’t tackled the roundups because there’s just so much. And with some I’ve ended up making copies for another site, which is needlessly redundant. Couldn’t I just make one main version right here, in a new repository or wherever, and code in some subject toggles & transcribe as I go along?

Anyway, could I recode a roundup from the ground up? Probably. Although right now I’m still struggling to find a good compromise between integrating images with typed text vs keeping things responsive (latest attempt is either one-column responsive or full non-responsive, with no in-between), which I’d definitely want to figure out beforehand… namely because

  1. even though handwriting all the notes takes me forever, I think it adds a fun personal element that gets completely obliterated with typed text but then I’m just writing and typing things, which kinda defeats the point; and
  2. sometimes when putting sketches together some funny juxtapositions happen, which can prompt doodles of their own, and that gets lost if everything turns into one column! even if two-column sucks on a tiny screen, which also deeats the point of bothering to make it mobile-friendly!!

Maybe the solution will just be to keep saying to hell with “mobile” designs and forcing everything to be the same size, with pixels. The handwriting thing will unfortunately be either lost or redundant regardless (unless I make a font of it or something, which… eh? not off the table but I would want to get more in-depth with it than what generators spit out, didn’t take a typography class for nothing), but at least there’d be some semblance of a linear document flow. Oh, and the toggles and potential zoom-ins and all that good stuff. That was the main thing.

speaking of site/art stuff

Describing images is a heck of a thing sometimes. Like, the image earlier in this post is straightforward—Screenshot of an unread Gmail message from Zell Liew, titled “Designing a focus style” and dated to October 2nd, 2019. Utilitarian. To-the-point. That’s all it needs to be. If the URL broke, nothing significant would be lost.

Inevitably, though, accessibility calls for image descriptions, and that means art descriptions, and like… I don’t know, you could be utilitarian with it. “Woman smiling mysteriously” or “Painted portrait of a woman with a mysterious smile, on a natural background” or “[insert in-depth physical description of the woman to the point where it starts to sound mildly creepy, the kind of indulgent trait-listing a novelist would gawk at, never mind the rest of the details]” is objectively correct, but they don’t convey the essence of the thing.

And sure, picture’s worth a thousand words, even utilitarian images can be captioned differently depending how they’re used.

But the point of a visual is that it can convey an entire idea in an instant—the drama of chiaroscuro, the humbling of a downshot. Inherently, reading cannot do this. Prose is, well, wordy; poetry gets close but it’s not the same.

Of course, it doesn’t have to be the same. It’s just… something something essence, it gets me thinking about the nature of a piece (of writing, of visual art) and how many ways the same exact idea can be conveyed, and also I just found this “Alt Text as Poetry” project while searching for something to link for the “captions can be different” text above, and that site looks extremely relevant to my interests and I lost my train of thought a bit.

Oh, and comics as a medium already deal with text/image integration, but potentially also run into a similar “losing something if it’s a font” problem as the aforementioned colorscript thing, and. Yeah.

Art.

one more philosophy-adjacent art thign

Main reason I’m writing a blog post at all today is actually to save a comment I just left on a video, because I don’t know if I’ll ever get around to writing a Comprehensive Post™ about the topic but it is interesting and YouTube comment sections are an ephemeral wasteland.

Run-on aside.

Brief context: Discovered Royal Skies during recent dives into Blender hell, and as an unsolicited + unsponsored side note those tutorials are extremely heckin great. I am typically loathe to watch videos when whatever they’re saying could’ve been an article, because time-based media is cool but for instructional purposes skimming is cooler. Video tutorials will almost inevitably be paused along the way, and the time adds up quick when there’s like, 30 minutes for one (1) task.

But anyway.

Blender tutorials aren’t the only thing on the channel, and lately the guy* has been making a series on AI art, currently with a focus on Dall-E. Some of these “tutorials” are short enough to get snatched up by the YouTube Shorts thing—read, under 30 seconds long. Because you’re typing in prompts, waiting for it the load, and bam, ya done.

I don’t really have a strong opinion on AI art, as a concept, right now. It’s a cool novelty, I guess? I don’t think it’s the death of the art world or other handwring-y sentiments, although (from my extremely cursory knowledge) there are extremely valid concerns about the algorithms reflecting human biases (what else is new) and machines replacing jobs (what, genuinely, else is new). We don’t have good answers to these things yet, but we truck ahead with the tech anyway, as we do. Hell, we don’t have a good “answer” to the car yet. But I digress.

Here’s the video I commented on, and here’s the comment:

Man, watching as a casual digital image-maker (usually draw, trying to get into 3d—hence finding this channel!), I think this stuff is fascinating. Started watching these as something of a morbid curiosity, initially not thinking of AI ~art (AI-generated imagery, whatever you want to call it) as more than a novelty, but there’s clearly some degree of skill involved in making it—even if the type of skill is more akin to an image search than physical dexterity. And I would argue there’s skill, or at least specialized knowledge, involved in database- or in this case algorithm-searching; for instance, finding references for odd poses or objects can be a huge time-suck—and even outside the realm of images, compare “how do i become an artist quickly” to “beginner drawing tutorial fundamentals [with a time filter for, say, under 10 minutes]”.

Like, look at all the words in the final query here; you need to do some research to know the camera specifics (30mm?), why it’s important to specify it… I can discern why the “trending on Nike website” part might be important, because I know a bit about commercial web design; web designers need to account for mobile devices, so the images need to be even bolder and simpler than even a poster (size is set, can be huge) or commercial (implies video, things can move in/out of focus). But I have no clue why 30mm is the camera setting of choice! You’d need, I assume, at least cursory know-how of advertising photography for that.

Anyway, especially after reading some of the comment backlash (which has more to do with the mere concept of AI than anything actually being presented in a How To Use AI Tutorial—valid concerns, but imo kiiind of besides the point?), just wanted to pop in and say it’s really cool to see how this stuff actually works. You’ve definitely got me reconsidering my initial dismissal of AI-generated imagery and the ~nature of art~ and all that philosophical stuff, even if that’s also besides the point :P Genuinely curious to see what the rest of the series has in store, and thanks for making such concise tutorials in general!

(N.b., I rarely log in to YouTube. And I am definitely not getting into comment fights. It’s an interesting topic worth discussion, though, so feel free to reply there and/or hit me up through some other way. I’ll, uh, get to it when I get to it. probably not in 3 years.)

Anywho, something I didn’t mention in the comment (because it would be so so very besides the point that all my parentheses and asides would become a self-consuming black hole of tangents) is that I was reminded of my image-describing struggle. Seriously, look at this:

Screenshot of a diagram in the video; context described below.

A “prompt engineer” uses roughly nine attributes for best image-generation results: mood/emotion, quality, camera lens, source (where the image might be found), description, subject, setting, purpose, and destination. I think that shit’s fascinating!!

Of course, that’s how the AI “gets” it. And a human reader can understand the patterns being matched here, more or less—although, as mentioned (e.g. the camera lens thing), not necessarily why.

And for art, maybe some of these details just aren’t important? Narrative, I’m thinking narrative; “city sidewalk at night, Dutch angle” doesn’t mean anything if you don’t know what a Dutch angle is, and even then, what do you know except “conveys tension”? It’s a technique, and the why is more important than execution. Are Dutch angles used at key scenes to hint at the presence of a specific character, or did the artist/cinematographer just want to sprinkle in some drama? “City sidewalk at night, Dutch angle” ≠ “Night. Crooked city; keep your head down.” The latter evokes an image, even if it’s not the exact same as the one that’s been drawn.

I rewrote that last paragraph several times. Figuratively I am banging my fists on the table in a repeated gesture of undirected enthusiasm. Art talk! Art in general!! Describing art, the way words are objective and yet not—frustrating, fascinating, I love it and I hate it and that’s why I write walls like this!!! God!!!!

P.S. on the AI thing

Again, no strong opinion on AI art as a concept (ignoring socioeconomic factors, or at least trying to, for the time), whether it can “truly” be considered “art.” But a lot of arguments against it— I mean, I haven’t done a deep dive, but the aforementioned comment backlash and similar sentiments… It Don’t Sit Right. So here’s some more Thoughts™, in a place that will not get me immediately dogpiled:

Arguably one of the biggest elephants (AIlephants?) in the room is the idea that AI art is “stolen,” because it’s been trained on All Of The Images, regardless of whether the image creators gave permission for it. (Boy, copyright ethicists are gonna have a swell time if we ever manage to bring people back from the dead. Starry Night print, anyone?)

Tongue-in-cheek wild hypotheticals aside, though… Well, Royal Skies already made a video on it, and he argues that it’s not theft because the AI is pattern-matching & creating something entirely original based on those patterns; there’s no copy-pasting going on here. It’s not even collaging in the “throw a bunch of fruits in a blender and let it rip” sense, it’s more like watching 5,000 fruit smoothie tutorials and then winging it based on what you keep seeing people use.

Personally, I found that pretty convincing. But also, I just have a very lax approach to how art is “owned” on the internet in general. (No, I’m not talking about NFTs; I think those are silly (and no help to the environment, to boot).*) Look, I grew up with “original character do not steal” descriptions and copyright disclaimers in fanfic and giant watermarks plastered everywhere (I still see giant watermarks!), and if there’s one thing that the internet has made very clear it’s that it is a very public space. Anyone can and will use anything on it, at any time, for better and for worse and for everything in-between.

Also, more generally, I think that everything is derivative. The most “original” ideas are still synthesized from previously-experienced work; nothing is invented in a vacuum, not common objects and definitely not the kinds of creative work people would typically call “art.” (As in, you could definitely argue there’s an art to product design. But most people’s definition of “art” includes, say, paintings. And guess what? Both derivative!)

That all being said: You realistically cannot prevent your work from being scanned by the AIgorithm unless you never share it publicly anywhere ever, and even then who’s to stop the rogue reposter. (Which is an issue in its own right but I’ve digressed how many times now? Yeah, maybe another post.) At any rate, I don’t think artwork being scanned by the pattern-detecting gods should be any more of an issue than a real human being taking stylistic inspiration from a favorite artist. The derivative thing is a machine this time, not a human, but… imo, so what?

And, yes, I acknowledge that “style theft” is something some artists push back against anyway. Heck if I know what’s going through their heads, frankly. Personally, as long as there’s no money being made off copying some indie rando, I got no issue with it.

TL;DR AI isn’t “stealing” anything, objectively, and on a concept level I got no beef. If I knew more about how any of this works I’d be tempted to train an AI on my own art, frankly; I think that would be fucking hilarious.

Section TL;DR: I don’t know if I love AI-generated imagery but I don’t think it’s a bogeyman, and don’t buy the arguments I’ve seen about it being “stolen” or “not art.”

Also, on static blog posts, no one can pick an internet fight with you (directly, anyway).

one very last thing for this monster of a post

One of these days I want to make a follow-up to the “Tale of Two Projects” post, but the most pertinent things I don’t want to forget to note:

Okay phew holy shit I think I’m done here, 147 code-lines and counting. I guess this could have been separate posts after all, but I still have this nagging idea that Posts About Particular Things need to be all cleaned-up and essay like. Using my blog as a sporadic but relatively-unfiltered public journal is clearly a better idea.

I’m, uh, workin’ on it. Thanks for reading if you made it this far, have a virtual fistbump. 👊