Code wizards have been using machine-learning tools such as VQGAN + CLIP and BigSleep to create novel images that grow spontaneously from word prompts.
To my fascination and delight, some of them have gotten interesting results by including the phrase "by James Gurney."
'The Hermit Alchemist’s Hut by Van Gogh'
Erfurt Latrine Disaster (Twitter @ErfurtLatrine) Prompt: "Towers" #VQGAN+#CLIP
The prompts can be simple, such as "Towers."
....Or the prompts can evoke a particular role-playing game, such as "Forbidden Lands."
The results develop a unusual style if you add a descriptor naming a studio, portfolio website, or rendering software, such as "from Studio Ghibli" or "trending on ArtStation" or "rendered in Unreal Engine"
"The Grand Hall of the Sacred Library by James Gurney"
Twitter user Ryan Moulton @moultano created a set of related images starting with the phrase 'The Hermit Alchemist’s and varying only the style cue:
'The Hermit Alchemist’s Hut by James Gurney'
'The Hermit Alchemist’s Hut rendered in Unreal Engine'.
'The Hermit Alchemist’s Hut by Van Gogh'
Matteo Vera @MatteoVera1 VQGAN + CLIP
"A castle built on the skeleton of a dead god by James Gurney"
Ryan Moulton @moultano "In the Woods, Gouache Painting."
Using the phrase "In the Woods + Gouache Painting" (without an artist's name) yields something that appears painted in water media, like a Mary Blair concept painting, but with something weird about the kids' faces.
All of the results have issues of basic logic and perspective. They never make sense or seem fully coherent, at least not yet.
But some of them do suggest a recognizable style. Does this look like my style to you? I'm not sure; it feels both familiar and alien. It almost looks like something from a long lost sketchbook.
---
Blog post by Ryan Moulton "Tour of the Sacred Library" using my name in the prompts.
12 comments:
No, not really, James. One or two seem vaguely/remotely like a few of your paintings, but overall I'd never name you as the artist imitated if I wasn't told.
Eugene, I wonder if the dataset they're drawing from includes the academic painters and illustrators that I've featured on my blog.
James (and Eugene),
Indeed, I am wondering along the same lines - what has gone into the initial assumptions (andthe dataset)? James, I'm curious about how much of your initial response is "wishful thinking". And I want to know what you, yourself, might have put into the initial dataset. I want to see your results - your own "self-portrait". So, are you encouraging us, by this posting, to get into the software and play around for ourselves?
James (and Eugene), right along the same lines, I'm also wondering about whether the initial conditions include the contents of all of your blog or just James' work. And James, is it way too complicated to think about doing just that -- to put in what you think ought to be the initial dataset? I agree with arenhaus, that much of my response is "wishful thinking." James, are you encouraging us to get into the software to munge around? It does seem intriguing - like a rabbit trail. We used to do this, starting about 20 years ago, putting in a well-known artist's work, example paintings, along with a photograph, and the result was a painting in the artist's style. What a time-sink, what a rabbit trail that was! and now, here we are!
JRS, to be clear, other people created these, not me——I'm just sharing. I understood the code jockeys to say that putting my name into the prompt just led to interesting results, not necessarily that it made the results look like my art style.
The thing about these machine learning algorithms that's mysterious and intriguing is that they can't tell you exactly what data they're using or how they're analyzing it and recombining it.
I’m not impressed.
You would think that out of four “…by James Gurney” prompts, at least one of them would show a dinosaur dressed in medieval garb, standing in a parking lot. -RQ
combining specific text prompts with CLIP, etc. has yielded some really interesting results. I sure appreciate you making these posts (although people seem to misinterpret). Learning to use this software is one of my goals.
I can see some underline of your style in the pictures. In any case, the results are very interesting and have very strange surreal feeling.
How very intriguing! Is it too early to beg for the "iPad app"? I'm thinking that an app like this might help on those days that I have "painter's block"...the days when that big expanse of prepared blank canvas just doesn't speak to me....the canvas just stares back, glumly, like an over-stressed adolescent, daring me to mar the surface, freezing me into place...sometimes, that feeling is just overwhelming...
You can do this on an iPad already, or even on a phone. The images are created by a bot on the Eleuther AI discord channel, which anyone can join using the link here: https://www.eleuther.ai/get-involved/
The bot is in the room called “the faraday cage” and all you have to do in there is type .imagine followed by your prompt. The room is full of all kinds of weird artworks that you can add to to your heart’s content - though each one takes five minutes to render!
Post a Comment