So I've been babbling a lot about generative art strategies recently. And people seem somewhat confused about what i'm talking about. So i'll try to provide some examples in this forum discussion. I'll be adding to this discussion over time to showcase different approaches.
What is an art strategy?
You can think of it as the 'recipe' to create a particular artistic style.
What i mean by this is the series of different steps to get from point A to point B.
Point A is a blank canvas (white in my particular case).
Point B is the particular style you are shooting for.
That could be many things, including
1: a particular paint or visual effect style
-natural media looks, like watercolor, oil, chalk, pencil, etc
- art history styles, like pop art, cubist, etc
- particular artist's style, like Francis Bacon, Chuck Close, Albrect Durer
- marking technology, like woodcut, engraving, screen print, etc
2: a mood, feel, emotion, etc
- calming, busy, evil, intense, etc
3: something different
- let's see where we get taken as we try new ideas
4: something that makes sense artistically to you
We could endlessly flesh this list out, but i'm hoping you get the idea
What is a generative art strategy?
One could create any art strategy by executing a series of steps necessary to create it.
A classic example would be what i will call 'generic make a painting approach #1'.
This generic approach to building a painting is very straightforward.
1: One starts with a fairly large brush size, and roughs in the background. Broad strokes of color, not paying too much attention to detail. It's really just about covering the canvas in an expedient manor.
2: You now start reducing the brush size, and paying more attention to detail.
3: Repeat step 2 until finished.
Now the classic beginner Studio Artist approach is to choose some random factory paint preset, hit action, let it draw, then complain that it's not really a finished painting (because there's too much uniformity).
I would answer that it's because you didn't follow the 3 steps above. Or some other more sophisticated art strategy.
Of course you could instead choose a Paint Action Sequence (PASeq), which is a series of Studio Artist actions. The whole point of PASeqs in Studio Artist is to encapsulate 'art strategies' into a single action script.
But the crux of the above discussion is that the artist has to do some work to build an art strategy. And people are oftentimes lazy. So they just press the button, and then are either happy, or complain it isn't doing what they want.
So, generative art strategies to the rescue.
Generative art strategies build the individual components of an art strategy for you.
In the penultimate example of a generative art strategy, you simply press the 'generate' button, and Studio Artist does all of this tedious 'build an art strategy from hand' stuff for you.
Of course Studio Artist is really all about 'auto-intelligent assisted' creativity. So the artist can be directly inside the active intelligence loop. Directing it, controlling it. This is a thousand times more interesting than systems that just do it all, and then you are stuck with the end product. It's also how you as an individual artist with individual feelings about what is good or bad about art to be a part of the creative process.
Now there are many different ways to build many different art strategies. And many different ways to 'generate' art strategies.
And artists have aesthetic opinions.
So some approaches may resonate with you. Some may not. That's not a problem, that's called having an opinion about what you aesthetically like or don't like.
And Studio Artist tries to address everyone's needs (for the most part). Without having an opinion (for the most part). Because everyone at Synthetik Software learned very early on that one artist's greatest effect ever is another artist's worst piece of crap effect ever. So we give you many, many, many different choices, and let you decide what makes sense for you to gravitate towards.
So i'm going to be presenting a lot of different approaches to this whole art strategy and generative art strategy thing in various posts in this thread (and elsewhere). Some of them will resonate with you, others might turn you off. Try to keep an open mind.
If there's any real these to the Studio Artist V5.5 new features, many try to begin to address what i like to refer to as 'visual improvisation'. Features to let you be more free and spontaneous in your approach. maybe just to stimulate your thinking. Or to enable you to stumble onto something interesting. that you can then pursue in meticulous fine tune editing if you so desire.
And if you hate what comes up, enable you to quickly move on to the next thing to check out.
So let's start with an art strategy that is really about being spontaneous (as opposed to honing in on emulating some particular style, or technique).
So you goto a particular paint synthesizer category of interest. Or just choose one at random.
Hit the action button. Let it start painting. Now start mutating the preset in real time as it's painting.
I like to mutate the source images as well.
In SA V5.5 i can do both of these from the front of the main workspace.
In the new Sa V5.5 PowerToolBar, there is a GSrc button that let's you create new generative source images. The definition of that can just be a new random source image, or an algorithmically modified source image, or something totally procedural.
There is also a M button in the Preset Browser that let's you mutate the current preset on the fly.
So you fire it up and start pressing either of these 2 buttons. Until you get somewhere interesting.
Not my greatest painting by far. But it took about 30 seconds. And i got a new wet paint preset i liked i saved in a Favorties folder as well. To use later in a more controlled way.
Now at this point you may notice that i violated my 'generic make a painting approach #1' definition. I was not working with anything to build steps 2 and 3.
That's not totally true, because my preset mutations did create presets on the fly in my 'make something interesting happen' mutation session that focus attention. Focusing attention (by focusing on edges or texture for example) can be thought of as a different kind of step 2 and 3, for a different kind of art strategy.
But suppose i was interested in my original steps 1,2,3 art strategy.
I could of course manually create a PASeq that implement the steps in the art strategy. So start with a large brush, no real focus on details, then add subsequent paint actions that decrease the brush size, and also add additional edits to the preset that make it focus more (intelligent brush modulation, intelligent path end checks, etc).
And you have always been able to do this in Studio Artist from the very beginning V1 release. So why do so many people never get around to doing that?
That question has perplexed me from then until now. Which is why we are very focused on these new generative art strategy features. Because we're trying to make it as easy as pressing a button.
So let's do that. Let's create a simple variation on my 'generic make a painting approach #1' art strategy definition. I will use the GPSRun button in the new SA V5.5 PowerToolBar. And i'm going to use my new Favorites paint preset i created in my random preset mutation session above.
And i get this:
To really see what the generative art strategy is doing, let's grab a different factory paint preset. The classic 'A Gradient Brush' preset from V4.
If i just hit action for that preset, i get the following.
If i use the GPSRun button in the new SA V5.5 PowerToolBar, i get the following.
All the above example have (in my opinion ;-) a basic flaw: they all rely on an underlying form totally taken for granted, not unearthed or discovered, certainly not created.
This is what I have called years ago “Visual Karaoke”.
I know that most people will be happy to work with/from that kind of structure, but there are some who want the underlying form to be the reward of the work, not its taken for granted starting point.
As you know, I have been confined to working with computers for many many years, natural media could kill me, but I now have enough experience to claim that, even with computers, one can work with those tools and avoid being locked-in from the start by/in whatever structure the source will provide (unless of course that is what one wants).
This is said in a spirit of coontribution to the thread, I have no intention to criticize here.
I understand where you are coming from. And sure, they are 'visual recipes' to a certain extent. Useful for many people, but not what you are shooting for in your work.
We should try to figure out what 'generative strategies' make sense for you. Because you can build your own. And certainly that's where i hope people will end up. But i also know many people don't want to put the time in. So the new features in V5.5 let them have fun and be creative without putting in the design work.
I do think the new visual improvisation features in SA V5.5 will be of interest to you once you start playing with them. At least i hope so.
Looking forward to SA V5.5 now that I have an iMac running the appropriate macOS and with enough power to hopefully make the new SA sing.
I’ve done enough teaching, lectures and masterclasses to trust there always will be some people as weird as I am to want to probe the stuff I am talking about.
Even if SA provides “visual recipes”, it has been a major contributor to the work I have been able to do for nearly 20 years, and I am allergic to recipes. ;-)
And I am very grateful to you for having made much of that work possible.
I'll think about trying to define some of the strategies SA could use in making generative art which would not be trapped in the source's underlying form.
But now I have to go back to the kitchen, am the one cooking the New Year Eve meal here!
Have a Happy One John, we’ll be in 2021 several hours before you!
Your comment is valid for the examples posted above so far (ie. that they rely on an underlying form that was not created). But it's pretty easy to setup a different generative art strategy (either manual or automatic) where that is not the case.
Almost everything i was generating using gallery show last night was not what you would consider 'visual karaoke', which i'm taking to mean building some 'art' output that is directly generated based off of the current loaded source image.
My intention with this loose tutorial on generative art strategies was to start simple, and then get more complex. Because if you don't understand the simple cases, you probably are not going to understand the complex cases.
Recursive generative art strategies being a great example of something more complex that can be targeted to be representational, or to create structure that does not exist in the source image you are working with.
I'll try to move things in a direction you might be more interested in in future posts.
For one thing, most people start with a structure they will be tied to for the length of the genesis of the piece.
That is already one hell of a limitation, but when that structure is not even of their own making, that is what I mean by visual karaoke.
One only needs to view great painters at work to realize that they are in search of an elusive structure, they do not rely on it.
And in that search, there are at least 2 poles: one is akin to “Where is it?” and the other “Aha! There it is!”
It also manifests itself through “From the outside in” and “From the inside out”.
Or/and from “Stasis-reclaiming” to “Flux-reclaiming” (when things change too much, we get scared, we make a “reclamation" for stability/stasis, and when things remain too much the same, we get bored, and we make a reclamation for change/flux).
Those swings are constant, few people are aware of what goes on "inside” them, and often lock onto one pole or the other, acting as if it were the full picture.
This is a given if working honestly, sincerely, not working towards a product, not making trinkets, but being deeply involved in a process which can teach us a lot about ourselves, about our self.
The constant swing from one pole to the other, if one surrenders to it wilfully, with lots of good will, leads to a failure of sorts, a bit like the ashes left behind when the fire has died.
But, in Albert Camus’ words, “The failure shall be the measure of success”, and it is precisely the traces left by that constant swinging, those ashes, which create, no, are, the real work of art.
Which brings me back to possibly the most important “thing” I have learned in nearly 60 years lived as a painter (first in natural media, now digital): the best and possibly only way for avoiding making trinkets is to work “by-way-of-not-knowing”.
But, “my” way of not knowing, not somebody else’s.
That is the closest to a real strategy I can come to.
I have to be honest, the intention of this post was to show off some new Studio Artist V5.5 features. And to elaborate on the whole notion of 'generative strategies' or 'process', which i think having some understanding of is essential to understanding how to take advantage of some of the new SA V5.5 features.
To some extent it seems like you don't want me to explain to people how to use the program? And i really feel like i'm being pulled off task from just explaining to people how to do things.
I'm also somewhat mentally ragged and caffeine jagged from working ridiculous hours to finish up V5.5, so i'm going to apologize preemptively for being crabby.
The notion of the 'source' is really an essential feature of Studio Artist. It has been from day one. You can read all about it in the fabulous User Guide.
It's always there, even if you setup the source area to use the color picker. It's always modulating things internally to the paint you are working with. Yes, you could actively hunt down and turn off all of that internal modulation, and then never use anything but custom manual paint with no auto-draw interactivity. But i don't know of anyone who works that way.
So, it's a given that the 'source' is being used to construct 'visual attributes' based on human perceptual modeling. And those visual attributes are made available over over the program (not just paint, but image processing effects, etc). To actively modulate all kinds of different things the program is doing.
So there is no escaping working with the 'source' if you are using Studio Artist.
Now there are as many different ways to use the source area as there are artists.
You can use it in a totally traditional representational manor. Take a picture of a bowl of fruit on a table, and then bask in the glow of painting a traditional still life. Or bask in the glow of pressing the Action button and watching Studio Artist do it for you.
And i'm fine with people working that way if they want to.
But at the same time, i'm trying to push people to think outside of the conventional box.
How can we expand our definition of what the 'source means'.
One art strategy that i really like is to first 'construct your own custom source'.
What does that mean?
It can mean a lot of things. Suppose you want to create a painting of some figure (let's say political for the sake of this discussion), but you'd like it to be a gross caricature of that person's face. With grossly exaggerated features, and maybe a reptile head in there somehow (because you're trying to make a statement in the piece).
Great. So construct this mythical desired source image in the Studio Artist canvas first. Use interactive warp effects to do the caricature modifications to the facial structure. Use your reptile image brush to get that donkey head in there somewhere.
Work with the contrast and edge structure of your source image. Because you want to modify it to encourage the later painting of it to have the characteristics you want. And you can always make it better, better suited to doing that.
Now run Canvas : Canvas to : Source menu command.
And now you have a custom designed source image you can use to build up a representation painting of a mythical figure you constructed.
Now if you studied and practiced manual drawing and caricature for years, you could just skip this step and create the finished product from scratch in the canvas, But we want to let interested people create visual things today (we don't have to call it art if we don't want to, we can call it a constructed visual artifact) without having to take years to practice and develop hand eye motor coordination reflexes.
Of course there are as many different variations of this simple 'make a custom modified source to your liking' art strategy as there are artists.
One fun one is to use something like the Vectorizer to create an abstraction of a photograph. And then use that as a source starting point for a painting or other visual effect. Gallery Show can do this for you automatically.
But no one said you were limited to working with a single source. In my work with Studio Artist i typically work with multiple curated source images stuck in a folder.
What does curated mean?
It means you chose them for some reason. Could be as simple as you like them. Or for their color, or textural properties, or political significance, or mood, or use your imagination.
And of course you are not limited to a single folder of images, because if you are working with image folder brushes or image folder background textures, you could be working with multiple sets of different images at the same time.
Now there is always an element of chance going on in the background somewhere when using Studio Artist. Random and procedural random and image modulated random all jiggling parameter values in the background.
And when you are working with a collection of different source images as your 'source', you are like that lost soul in 'The Man in the High Castle' throwing the Yi Jing bamboo sticks to see into the future. Because the particular image you are using at any given moment in the 'process' of creating your painted canvas is always changing.
The real 'source' you are working with is some amalgamation of the visual characteristics of the individual images inside of that collection you are working with. One you curated for it's particular properties.
You can think of your 'source' as a database of visual attributes in this particular generative art strategy. Again, if you are stuck on the word 'art' because you think that word means something specific that this whole thing isn't then ignore it and use generative visual construction strategy instead.
But we're still not done on this I Ching 'what is the nature of a source' path. Because when your generative art strategy is a recursive cycling process with a feedback element internal to the 'process', the whole notion of what the 'source' even is becomes very ambiguous. Because your process might be creating 'reaction-diffusion' textures and shapes out of thin air, and whatever happens to be in the source area at any given time is like the ridges in a pile of sand that influence the direction a grain of sand takes as it rolls down the pile.
So there's a quick run through (brain dump) on some different approaches one could take to working with the 'source' in Studio Artist. Don't be afraid to explore them (and others you think up). because again, there are as many ways to do this as there are artists (excuse me, visual artifact constructors).
More Studio Artist V5.5 testing as we hunt down and kill those nasty bugs (and refine features based on working with them). So here's an example doing the exact same thing i discussed above, but using a generative source image. No real world images were used to create this painting, but the technique and generative strategy are pretty much exactly the same as my initial example above.
You can create generative source images in Studio Artist V5.5 using the GSrc button in the PowerTool Bar. The source image was generated procedurally in this example, so it's based on a unique mathematical abstraction created when the GSrc button was pressed.
The generative strategy is very similar to the initial example in this post on generative paint strategies as an alternative to random or constrained randomization for creating new presets. So i was working with generative dynamic brush effects using the GPSRun button in the new Studio Artist V5.5 PowerTool Bar.
I was also using another new generative strategy feature in SA V5.5 that lets you create randomized but still constrained designed adaptive selection masks on the fly. So i press the GSelMask button in the PowerTool Bar to improvise a selection mask for painting. If i like what came up, i then press the GPSRun button to create a new generative dynamic brush paint effect that is painted with the dynamic mask.
I pressed the GSrc button several times during my cycles through the overall generative strategy. So i was working with several different unique procedural source images to generate the final painting.
I should point out that this whole generative source feature idea is really just getting started. Once we add GANs into the equation at some future point in our Studio Artist development cycle, the whole notion of the 'source image' becomes even more fluid and dynamic.
Is a source image you created from interpolating within a latent space you created by training a GAN neural network on a database of real world images still 'representational'. It's a unique image that does not exist in the real world. It was artificially created by learning a set of perceptual feature vectors derived from analysis of a collection of real world images. It might look like something real, but it is not.
Fascinating questions. Fascinating things to look forward to.
I have to honest I was only tangentially aware of GANs until your post in late November and have since gone down the rabbit hole. The unique piece here in StudioArtist that I feel is going to be a lot of fun to dig into is setting source, style ( possibly created organically on the fly) and if I fully followed your post a sense of training it as well. This is going to be fun, because from what I can tell from the other GANs based tools out there you only select source and choose from one of their preset styles.
Sounds like a lot of thought has gone into this and I am excited to start my own Art Strategy. :)
There are many levels, models, of perception, the one I am most fascinated by is “first-person perception”.
That is the one that acknowledges the central position of the artist in (the genesis of) any art work.
So many people, including “artists”, have fallen for the "fallacy of misplaced concreteness”, they are not aware of the enormous differences between what they see, and what they think they see.
And that difference is at the very heart of Art, at least if one considers Art to be more than making trinkets.
One thing that is so often lacking in the automated processes, is the sensitive distribution of attention, a distribution of attention that is made of/by one’s awareness of the all essential “sense-giving” and “sense-receiving”, an awareness that creates/builds/constitues the very placement of the marks on the surface (be it canvas or screen), thus creating a tension, a dialogue, an Art work, which is sorely missing in most pieces done by automated processes.
Don’t get me wrong, it is not the automated processes which are (necessarily) at fault (hell, I have been using SA for almost 20 years), it is the way they are being used.
I do not have the time to really explore this here in the depth it deserves, but here are a few things:
first an excerpt of the draft of an article I am still working on:
second, this is a flyer presenting a workshop I was to conduct at Western University (London, Ontario), with a group of phenomenologists (first-person perception, remember?) coming from as many as 67 different countries.
The symposium, of which the workshop was to be part, has been cancelled due to the pandemic, was moved to be held at the University of Alberta in Edmonton later this year, than cancelled again for the same reason, no clue as to when or if it will finally take place.
But here’s the flyer:
The awareness of the dialogue between sense-giving and sense-receiving when working gives birth to a quality of placement of the visual elements I have yet to see realized with any naive use of automated actions.
Here’s a prime example of what I mean by "placement” (one of Joan Mitchell's glorious pieces):
Finally, here's a link to a conversation between two old geezers, the one on the right is Charles Cajori, an old friend, I had a share of his loft on W 14 Street in NY when I started teaching at the Studio School in 1978 (Charles is dead now, more about his work here). Some of what Charles says around 04:10 and 05:40 is priceless, how “this” relates to “that”, and how any change made elsewhere makes the original relation of “this” to “that” no longer relevant.
I have no idea of how, or even if, you could incorporate some of those “things” in the way SA works, but I firmly state here that any mechanical model of perception is a fallacy (which is why many scientists working in cognitive science have now finally begun to look at Phenomenology), we are first and foremost subjective beings, the eye does not work like a camera and we do not all see the same thing.
And I haven't even touched on how our perception of the/our/my “here and now" is (also) shaped by memories, anticipation and the likes, it is a total mystery, including to oneself.
Food for thought for sure.
But did you ever consider that the human mind might be a mechanical model. And that consciousness is essentially making up excuses for thoughts and behavior that have already been determined by the underlying system. I'd say this statement is based on observations in modern neurobiology (which is true), but you would criticize it for that very reason, so i'll point out that Buddhists say the same thing as something they have figured out from internal observation.
And i'm not sure what 'camera' means these days. Certainly not what it did earlier in my lifetime.
The neural modeling in Studio Artist associated with visual attributes goes way beyond the eye, all the way through the visual system to the IT cortex. So it's already no longer tethered to what the traditional camera world view, and is rooted in subjective consciousness to some extent.
An the very fact that GANs (generative adversarial neural networks) work at all tells us something very important about human perception. That we can encode the very nature of human perception into a mechanical model inside of a neural net (at least the human perception of what is physically in an image, i don't think people have tried training them on mood or raw emotion, so something we can look forward to i guess).
I could wax at length on, but will save for another time, and another forum thread.
Anyway, all of this is extremely off topic from the nuts and bolts of how one can approach using new Studio Artist V5.5 features.
I would suggest that you start a separate thread when you get some time. In fact i would actively encourage you to do so. Then we can talk about this stuff at length over time without it clogging up a thread on how to use things like gallery show and the new PowerToolBar in SA V5.5. Everything you are saying is great to think about, and helps expand people's world view.
I'm happy to try and incorporate the things you are talking about into the underlying zeitgeist of Studio Artist. You just have to quantify them enough so that is even possible. Or people can riff on what that might entail after reading more about what you have to say on this topic.
One obvious thing to think about based on what you just said is the notion of 'placement'. I'd love to get a description of what you are referring to in enough detail so that it can be coded up and dropped into Studio Artist somewhere.
Going all the way back to your "What is a generative art strategy?" three step process at the start of this thread. I almost wish there was a game like MIPmap (video game graphics lingo) like option available to do what you describe. Like having a source preprocessed to multiple resolutions that could be sampled and swapped on the fly to achieve either rough or detailed surface effects.
However. I am more or less doing the opposite of starting rough and sampling detail when engaging with automated processes in SA.
I am not particularly interested in the image processing aspect of SA - but I occasional like to see what happens when I click the Action button. I am looking almost exclusively for Action that doesn't (just) abstract an image. Rather, I want to sample an image for some feature - like color or possibly shapes or "energy" - implied weight or motion. I am most interested when the result of hitting Action results in something that does NOT look like the source. I actually crave to escape the source as representing something and see an output that instead becomes its own independent visual experience/expression.
Some very simple examples attached. If you didn't see the source - you would not be able to identify what the output images came from.
While fiddling with an image and expecting to run some Action or auto brush on it I find I want to sample an image source for something I like about it. But also get very specific about what I want in the source image.
I find myself wanting to stop an action (if in progress) to be able to sample the canvas for a section I might like and send that sample to the image source. I have been doing this by literally stopping an Action Process or waiting for the Action to finish and then screen grabbing a section of the processed image - at any size - and dropping the screen grab into the source to run more Action.
This more or less describes a process I would like to have directly in SA. To be able to sample a section of a canvas image and replace the source image with the sample.
Maybe there is something like this available already? After all these years I am still very unschooled in what SA can do with automatic image processing.
Sampling the color. Hand drawing with some auto based Paint Synth preset:
Sampling the colors and masses. Action button.
This kind of split from the source as something to imitate/abstract is exciting for me. I really only look for a couple passes of whatever auto function is run - to get something.
Any extra tools that might be in the SA or added to it - like sampling from the canvas sending sample to source would be great.