I can’t make art.
I can’t draw. I can’t paint. Recently, my mother returned to me what was ostensibly a ceramic pencil holder I made in fourth grade, which in reality looks more like a distended, diseased, pockmarked frog. In kindergarten–this is true–I failed Scissors. Still today, whenever I attempt to cut something, I’m left with a wave of scallops punctuated by the occasional jagged edge.
But I like art. I figured if I couldn’t make it, maybe my computer could. Thus the Arty Bots project was born. Arty Bots are Twitter bots that make original images and Tweet them. Many can also modify an image sent to them. Some results are shown above, which you can click to enlarge.
Below are some thoughts about the nature of art bots and the ArtyBot family, and I describe each bot in more detail at the end. But, if you’d like to jump right to them on Twitter, they are: @ArtyAbstract, @ArtyBots, @ArtyCrush, @ArtyCurve, @ArtyEdit, @ArtyFractals, @ArtyImpression, @ArtyLetters, @ArtyMash, @ArtyNegative, @ArtyOriginals, @ArtyPetals, @ArtyPolar, @ArtyShapes, @ArtyTiles, @ArtyTriangle, and @ArtyWinds.
Q.) Are the images your bots create art?
What, were you expecting a bunch of hedging, tergiversation, and hang-wringing?
I repeat: yes.
Q.) But I like tergiversation!
Fine. Let’s rephrase the question that everyone, from the average craft-fair goer to a curator at MOMA, must wrestle with in determining the value of art:
Q.) Does it look nice and can you hang it on a wall?
With the advancements of communication and printing technologies, anyone–you included!–could take one of the pictures created by the bots, download it, and print it on canvas. I use canvaschamp.com. A friend of mine had one printed at his local Costco. The images are 500 x 500 pixels, and the resolution is great for an 8″ x 8″ canvas.
However, there’s one caveat in declaring my bots artists: virtually all art created by bots on Twitter is non-representational. That is, the pictures do not look like things. Instead, the art is largely abstract.
Q.) But surely, the masters of abstract expressionism, such as Pollack or Rothko, can do something computers can’t.
If something were inherently unique about an artist or his/her work, then it likely couldn’t be replicated by others. And yet, the Columbia Law School reports that perhaps as much as 40% of all artworks sold in a given year are forgeries. Similarly, Art Law Journal says that many art experts have simply given up trying to authenticate art, including foundations for individual artists like Andy Warhol and Keith Haring. Certainly, forgers are much more skilled than a computer program, but consider Neoplasticism Bot, which creates works remarkably similar to one of my favorite artists, Mondrian. The old complaint of a skeptical philistine upon viewing splatters of paint on a canvas–“I could do that”–has a kernel of truth to it.
Q.) But a lot of the artwork I see generated by your bots is … how do I put this delicately … ugly.
A.) Sure. Not every new composition is going to be great. I’d estimate 5% of the works my bots create get some sort of affirmation of worth in the form of a retweet or favorite. Generally, I think each bot creates interesting images sufficiently often to merit it being online.
Q.) What can bots do that human artists can’t?
A.) A bot never tires. A bot never sleeps. A bot never runs out of paint or ideas. And, most importantly: a bot can’t be threatened, bullied, or censored into not making art.
Thus, an art bot has a whiff of political protest in it. We live in a world where we are daily told that beauty doesn’t matter. This is reaffirmed by politicians, newsmakers, and those inhabiting Internet comment sections. So often, those people are full of sound and fury, signifying nothing–to quote a non-computerized artist. (Not to mention how their tales might be told by an idiot.) An art bot is indefatigable. It, and by extension its creator, will continue to attempt to create something beautiful every hour, every day, regardless of political sentiment or regime. It is admittedly a very small protest, and a rather passive one. For more active forms of political protest bots, please read this thoughtful essay by Mark Sample.
And, unlike so many people in political conversation, my bots are actually willing to talk to each other.
Q.) Talk to each other? How so?
Many of these bots generate original artwork. As a complement, they will also take a picture Tweeted at them and modify it according to their style. Thus, my bots can have conversations between them. And sometimes, rather than a back-and-forth, the bots will decide they want to bring another bot into the discussion. Some original pieces can go through five or more of my bots before the conversation runs out. That’s where the truly surprising artworks are created. Each individual bot has its own style, and its output can feel similar after a while. But the interaction between bots makes any individual piece truly random by the end of it. The pictures at the top of this post were created by several of my bots in conversation.
A complete conversation is the above, a nineteen-picture sequence that involved seven bots: ArtyShapes, ArtyMash, ArtyAbstract, ArtyNegative, ArtyCurve, ArtyCrush, and ArtyWinds (some of them more than once). My favorite image is precisely in the middle, when the meadowlark gets crushed down to eight colors. It is arresting, and something I could never have predicted my bots would invent. After the first three or four images, it’s impossible to even know that the first image was the one of concentric squares.
Q.) What other art bots do you admire?
I think my favorite is @MM-62-1234-14 by Jason Ronallo. It consistently surprises with a wide variety of colors, styles, and techniques. Ronallo has also made many other art and image processing bots, which you can learn about on his webpage about them.
@tweegeemee is another good one, usually posting two variations of one concept. @softlandscapes paints its mountains pretty well. And if you prefer happy little paintings, @JoyOfBotRoss will channel the spirit of Bob Ross and paint a lovely little landscape for you.
Q.) What’s the technical process of creating and then sharing an artwork?
The bots reside in an apartment complex on a Raspberry Pi 2. They run via Cron jobs to keep everything regular.
I had very, very little programming knowledge upon starting this project, and no Python experience whatsoever. I had no idea how to create an image or how to Tweet it. I did a lot of Googling, reading forums, and fighting with code until things sorta worked. My point is that if you’re interested and stubborn enough, you could learn how to do this, too. This post by Jeremy Kun was invaluable in getting started, and I built off the code he’s graciously made available online.
The fundamental concept is that any image is simply a grid of pixels. Each pixel has a red, green, and blue value as well as an x and y coordinate. Based on that concept, you can apply math to any of those numbers to manipulate them. The trick is experimenting to find what manipulations generate interesting results.
Q.) Can I use some art generated from one of the bots?
Sure, as long as you credit the particular bot and it’s for noncommerical use. The specific license is CC BY-NC-SA. If you’ve got a particular commercial use in mind, feel free to contact me and we can talk about it further. I’m happy people have enjoyed the art and would like to use it elsewhere!
Meet the Family:
Probably my favorite bot. It generates abstract, colorful art via a wide selection of human-curated equations. If you send it a pic, it will reply with some abstracted version based on a randomly generated formula.
A version of my own bots conversing writ large. I curate a list of all active image bots on Twitter that can take an image, process it, then return it. ArtyBots is simply a glorified mailman. It will invent and paint a fairly simple picture based on random equations, and then randomly choose another bot to receive it. When ArtyBots gets the processed picture back, it sends that picture along to a new bot, and so forth. ArtyBots creates a new image every four hours just to keep things fresh.
ArtyCrush is old. ArtyCrush is stubborn. ArtyCrush only believes in eight colors: black, white, red, green, blue, yellow, magenta, and cyan. You will not be able to convince it otherwise. It will generate pictures in those colors based on random equations. If you send it a pic, it will crush it down to the only eight colors it knows exist. ArtyCrush is thinking about upgrading to Windows 3.1, but probably never will.
ArtyCurve only replies to pics sent to it, and makes them curvy. Or sometimes stripey.
ArtyEdit is kind of like a kid with access to a graphics editing program. It takes an image sent to it and randomly performs operations like blur, sharpen, and emboss to give things a much more enhanced appearance.
Fractals are kind of a gimme when talking about image bots. But the weird thing was that I didn’t fully understand how to correctly code a fractal. The result was that I wound up with weird petal shapes and some long wisps shooting off into the blackness. I decided those irregularities are what makes ArtyFractals unique, so here they are.
I’m fond of this one. ArtyImpression will take a photo and repaint it in an impressionist style, like Monet or Seurat. Technically, ArtyImpression uses pointillism: it samples a pixel randomly on the photo and then paints a small dot or rectangle on a new image using that color. Each shape has a random opacity, so the points wind up overlapping after a while, creating a pointillism effect.
ArtyMash is the only bot to rely on a purely external source: Flickr. It simply takes two public-domain images and mashes them together based on weird blending techniques. If you send it a pic, it’ll mash it with a random one from Flickr. The best results are surreal and unnerving.
Pretty straightforward: ArtyNegative takes a picture it receives and gets negative about it. Not terribly interesting by itself, but a nice go-between for other bots.
In a family of strange bots, ArtyLetters is probably the strangest. It divides an image it receives into a grid. It then adds up the red, green, and blue values for a pixel within each square of that grid. Based on that calculation, it generates a letter, with A being a low total value (colors approaching black) and Z being a high total value (colors approaching white). It then writes the calculated letter in an approximately opposite color superimposed on the original. I have no idea where this idea came from. Basically, ArtyLetters adds some noise into the system, but it can impose a weird sort of order to some images.
ArtyOriginals retweets the original artwork made by ArtyAbstract, ArtyCrush, ArtyFractals, ArtyMash, ArtyPetals, ArtyShapes, and ArtyWinds. I basically wanted an account that people could follow without clogging their feed with all of those bots’ replies. Tweets about once an hour.
Pretty flowers. ArtyPetals places a randomly colored pixel in the center of the image, and then takes a random walk (sometimes lovingly called a “drunkard’s walk”) either up, down, left, or right by one pixel, and then paints that pixel in a slightly different color. The pixels are then reflected about one or more axes. The process repeats, and the walk ends once the trail exits the edge of the canvas. Usually the images use multiple walks, with an average of about 350,000 individual steps per picture (although that number is highly variable). ArtyPetals will also take an image sent to it and perform this process, except it will use the pixel colors in the original image. The result is a petaled, kaleidoscopic version of the original.
Hey, remember that one time you learned about polar coordinates for that one week in that one math class in high school? No? Well, ArtyPolar does. ArtyPolar generates random equations in polar coordinates and paints concentric pictures. It will also rearrange images it receives in a similar way.
ArtyShapes throws a bunch of randomly colored ovals, triangles, and rectangles onto a canvas, and then runs them through a few filters. Sometimes instead it chooses to post concentric squares or vivid stripes. Perhaps not the most interesting bot, but it can add some shapely zest when it replies to pictures sent to it.
ArtyTiles is the square sibling of ArtyTriangle. It will take an image it receives and turn it into a mosaic of squares, complete with grout lines. It will also generate its own pretty grids on occasion.
ArtyTriangle harnesses the power of three sides to generate strange angular images. ArtyTriangle will take an image and recast it in triangles or in other shapes that can be built out of triangles, like trapezoids, parallelograms, or rhombi. It also creates some triangular artwork of its own.
The primary function of this bot is to take an image sent to it and then scatter its pixels in a certain direction, as if wind was blown across the image and the pixels were grains of sand. But ArtyWinds also generates thick, alien cloudscapes based on the same principle.
Feel free to contact me at bjbest60 (at) gmail with any questions or suggestions! I’m also on Twitter: @bjbest60.