We have kind of slacked off on the monthly art challenges here recently. So i thought i'd set a new one up.
It's our desire to make this whole thing a little more interesting by turning it into an official contest with official rules and perhaps some official prizes. Those official rules associated with an official contest will hopefully be posted soon.
Until that happens, any discussion here refers to the intellectual challenge associated with this new Studio Artist Challenge. As opposed to any 'official contest' with 'official rules'.
Here's the challenge image.
The challenge solution will be able to start with some arbitrary image of a person with head and upper body visible in the source image. And end up with something that resembles the artistic style of the image above.
I'd prefer a Paint Action Sequence preset that pulls it off automatically.
If no one can successfully rise to that fully automatic solution to the challenge, i'll settle for a comprehensive tutorial on how to manually pull it off as a series of documented steps described in detail to the point where someone can follow them one by one to get to the end result. This would include the various presets you would use for each of the documented steps to get from the source image to the final stylized finished result.
There are a few different things going on in this particular challenge image. There's a bit of caricature going on. And of course the very distinctive cartoon style drawing for the sketch of the person. And a limited use of color fill to accent certain areas of the image.
Some hints to think about:
1: We've been experimenting recently with caricature here by using interactive warp on a source image, then loaded the warped image into the source area and doing painting or other processing off of that. the warp builds the caricature. then you derive your painting off of the warped caricature image.
There have been discussions here for using a folder of images to derive a generic face. Automatic caricature would then involve magnifying variations seen in your original source image to the standardized head image. So that's one approach to generating it automatically. If you are taking the manual description option to the solution, you need to describe how to use interactive warp to do it.
2: By clever use of Smooth followed by thresholding, you could possibly generate masked areas for the limited color fills seen in the challenge.
3: Thin vector pen paint is going to be key to the sketching part of it. You can use ip ops to generate vector paths that the paint synth can then paint in if you wish. Or you could use one of more ip ops if you are clever and do it that way.
4: Manual selection of the eye areas is allowed if you are building a PASeq. Or building your script for how to build it up with manual paint steps. V5 does automatic eye detection, but you don't have access to that, so we'll let you fake it by doing the eye selection manually.
5: I jut gave some hints for possible solutions, but don't let those limit your imagination. The best way to pull it off might be something totally different than my suggestions. I'm always surprised by the different approaches people come up with when working with Studio Artist.
SO, get to it people. It's doable, but it is going to stretch your brain to pull it off i believe.
We hope to have some exciting prizes associated with this intellectual challenge once it turns into an 'official contest'. More on that as it develops.
Replies
Paul Perlow and i were discussing this most recent Studio Artist Challenge via email. He also brought up how he thought the art of caricature was a very important part of what the original challenge illustration is all about, and what you need to figure out how to do well to successfully solve it.
He mentioned the work of another famous illustrator, David Levine, as another example of this kind of caricature illustration style. David's work is very interesting, but it does seem to have a different feel to me than Joseph Ciardiellos's work. David's line work is way more busy. While Joseph seems to get the point across with a remarkably small number of well placed pen strokes.
I think this gets at something else going on in this particular challenge image other than caricature i feel is really important, a very important and fascinating component of the overall artistic style it represents.
In addition to the caricature part, there is something going on in the really amazingly good pen illustration work where the artist is able to encode 3D perceptual information in how the 2D lines are arranged. Any artist who is really good at generating a successful sketch of a 3 dimensional face or scene with a few carefully drawn pen strokes on a flat 2 dimensional piece of paper is doing something quite remarkable. Because a viewer of that 2 dimensional artwork perceives it as being a 3 dimensional representation of that 3D face or object.
Why is that the case?
Obviously it has something to do with human visual perception. There is something associated with the human perception of 3 dimensional imagery that can be encoded in a few carefully drawn pen strokes on a flat piece of paper.
i find this fascinating, as do some other people like me, that try to straddle the boundaries of science, technology, and art.
There have actually been several technical papers published on this topic, that try to analyze specifically what a really good artist is doing to make this happen.
Part of it has to do with the neuroscience of visual perception. Because we can look at a flat 2D photo and perceive it's contents as having 3D structure.
We do that based on certain characteristics of that image. And how we interpret them visually in our brains. Specifically in the parts of our brain that process visual imagery.
A really good artist may not understand the neuroscience of how this works, but they have developed intuitive knowledge of it, because they have developed tricks or techniques that enable them to reproduce that key information inside of their 2 dimensional drawings. Using only a few very carefully drawn thin pen strokes.
And a successful solution to this particular Studio Artist Challenge requires solving this 'encoding' problem as well.
In fact, someone technically inclined might actually view this approach to drawing as a successful technical solution to compressing the original 3D image.
It's actually a hard problem for someone manually painting to solve really successfully.
And even more challenging to try and solve with some kind of fully automatic approach to generating a drawing or painting.
The discussion going on associated with this particular challenge (some of it off line in emails to me rather than directly posted here), may actually lead to a few small steps towards solving the automatic solution, or at least improving it. Some interesting things have already come up in these discussions, both online and off line.
And i really look forward to other attempts at solving this particular challenge by other Studio Artist users. Taking advantage of all of the features that Studio Artist has to offer.
Or any feedback from them on new features Studio Artist needs to provide to really pull it off successfully or easier, or whatever.
David Levines style is a combination of crazy blobby forms and a heavy coat of "modeling" with lines… Interesting and very "beefy" - not as light as the original challenge style (by any means)
The attached Paint Synth Presets were used to make the unplanned sketch here. All hand applied (drawn)… just winging a Lincoln-like profile. No reference image.
To really get the style - I would have to do a lot more preparing of a reference image and a lot more practicing with the line like tools I used.
The lines here were made with cd_ink_ballpoint01_alt02 for single lines hand applied
The hatch lines: cd_ink_ballpoint01_alt01_auto-tilthatch02
This preset responds to tilt and pressure. Tilt to direct the line angle and pressure to get the long and short lines.
I used both presets with the source color white to knock back some of the lines as well. Basically using the presets as erasers.
To do this look justice I would have to practice enough to lay down as few clean hatch lines as possible.
To make the job easier - I would make the drawing in two layers. One for the outlines. A second layer for the hatch lines. I would split it up like this to allow for some freer drawing of the hatch lines… That I could then cut (erase) to fit into (inside of) the outlines...
This image was all one layer - cleaning it up would be a pretty tedious task… with the lines intermixed as they are.
CD_lintest-pens.zip
Here is another shot at the original look and a combination of surface techniques from both the caricaturists discussed so far. This is again entirely hand drawn. I am happy with the way SA can be used to do an organic hand drawn style like this. It is worth presenting just as a counterpoint to automated possibilities. The fact that I can get this lively a look and enjoy the process has always been a selling point for me with SA.
(I will post some odd offshoots of trying to make a more automated set of Presets somewhere else or further along in this thread.)
The image was built from a fast sketch. This was done in in the close to defunct application TabletDraw - which is designed entirely for line type drawing and particularly good for sketching… So I am off the reservation with that part of the process.
However - for several reasons speed of drawing (faster the better for me) is a big factor in the very first stage of developing an idea. Often it is the action and sweep of the hand that expresses the line and form I want a thing to take. I am drawing as if passing my hands along the contours of the thing (an idea)… The speedier I can pass over the contours the more "organic" the results.
The two line and color samples show an approach that uses the multiple lines for shading:
lin01.jpg
and just contour lines and color.
lin02.jpg
The art was made with variations on previous Preset postings - attached with this post (zip file)
The images were built using three layers.
A color layer
A line layer (contour)
And a Multi line layer
The sketch was loaded in as a source image and I used the onion skin feature to view it in the canvas.
lintest_CDpaintsynth_presets.zip
On the original lighter line/contour drawing technique...
When I was a young learning artist I encountered several related ideas about how things can be represented (and I assume - interpreted) as having volume. Volume would be the word for three dimensional.
In Physiological psych classes - and basic biology for that matter - I was exposed to a lot of ideas about the mechanical structure of sight.
In art related classes i was exposed to Kimon Nicolaides techniques for learning to view things and represent aspects of them.
And
While I was staying with a sculptor friend of my family (many, many
years ago) - who was also very involved with techniques for helping the blind navigate their environment - we would occasionally have discussions about areas of the brain/eye that (while in the blind might seem dormant) are used to represent information. One of those areas was edge perception. Particularly the way the (working) eye processed edges. Deepening contrast at edges to "enhance" the "edginess". The edge of dark surfaces will have a slight halo of bright next to them - enhancing the edge. An effect that is viewer based rather than (perhaps) actually there. Artists use a similar edge enhancement technic when they paint border edges with a lighter color - a highlight stroke to pop out the edge line.
The sculptor I stayed with felt that even the blind must have wiring established to interpret edges - and might be utilizing that pathway to (mentally) construct things in a room from aural depth cues (echo ranging) as well as recollections of touching. They could be (essentially) eidetically creating a caricature of a room and things in it. With edges (front and back - volumes) integrated into the image.
They could describe things as "in front" and "behind" - why wouldn't they be seeing them (imaging) in a way similar to sighted people.
A technique like contour drawing (line drawing of edges) uses something of a similar trick - to enhance edge clarity. Its actually a double trick. The edge line (super enhanced in a line drawing - black on white) and overlap (front and back - volume). The simplest form of overlap can be illustrated with two circles. But can also be seen clearly in a quick ear study.
Wether a tool could be used to find edges and isolate overlap… Thats a big unknown. That overlaping edge question was one of the first things I noted in the original challenge art.
I would be fascinated with a tool that could treat fields of depth as if they were something like layers. Drawing a "background" image (clutter) onto one layer… Perhaps filling into unseen areas. Drawing foreground masses in to subsequent layers (on top)
Right now SA takes all information that is on the same level and treats it about equally. There is no front and back so to speak. Finding a way to layer information and automate the layering would be a fascinating trick.
Craig, you brought up 2 really important (and fascinating) points.
The first has to do with 3D perception of a thin line sketch being tied to where individual lines overlap, or alternatively where they meet, and then one of them vanishes.
This information provides valuable clues to someone looking at a 2D sketch about the 3D surfaces the sketch is trying to represent.
And they are also directly associated with similar 3D features that someone sees in a 3D object when they view it, which their brain then detects and uses to build the perception of depth in the view of the 3D image they experience.
Now the other fascinating thing you brought up is the concept of perceiving an image as a series of individual layers. Like a foreground object in front of a background. So the foreground object and the background it is in front of could be thought of in some sense as 2 different perceptual layers that your brain pulls apart and perceives as being different. This happens even when stereo vision is not directly occurring, which is the case when looking at a 2D photograph.
You are also right that naive use of Studio Artist to paint a given source image really makes no distinction between what could be perceived by a human viewer of that source photo as separate individual foreground and background layers.
There are however approaches one could use within Studio Artist today, to try and make distinctions between what the viewer might be perceiving as 2 very separate visual layers in the source image. Both with manual drawing, or when building PASeqs for fully automatic effects.
I do think it's important to distinguish what we are referring to here as a perceptual layer from the concept of individual canvas layers stacked on top of each other and composted together somehow. You could for example build up the perceptual layering effect we are referring to within a single canvas layer.
Clever use of selection is one approach. Keep in mind that how one can use selection in Studio Artist is way richer than in other more conventional graphics or paint programs. You can select an area of the canvas, and then fully mask painting or any other visual effects with that selection.
But you can also select an area of the canvas, and then use it as a part of your paint synthesizer programing, for example to force path start locations to only occur within the selected area, but to otherwise not mask any painting or drawing that takes place at all. This leads to a much more natural effect, than a hard mask of a paint stroke, which can look very contrived and artificial. Not the approach someone would use if they were actually drawing something, unless they were actually using a physical stencil to restrict their painting.
Representing positive and negative space is also an important stylistic technique for building the perception of multiple visual layers in a painted canvas. As well as helping to convey a sense of depth, or to distinguish foreground from background. So this is a 3rd new factor i'm introducing into the discussion here.
One could also speculate on additional neural processes that take place in the brain associated with the perception of positive and negative space in visual imagery, both real world, and in drawn simulations of that real world imagery. There are interesting scientific papers on this as well, in addition to edge perception, or perception of occlusion of intersecting lines.
The Smart Contrast ip op available within Studio Artist can be configured to derive very intelligent and fully automatic representations of positive and negative space in a source image it is processing. The fully automatic output of the Smart Contrast ip op could then be routed to the selection buffer for example, where it could be used to mask painting, either like a stencil, or in much more subtle ways, like i discussed above.
…….
Another whole take on this is that you could use 2 or more different source images at different times in an overall process to build up a final finished piece of artwork. There is actually a special feature in a PASeq called a Source Context you could use to build a PASeq that worked with multiple source images within the context of a single automatic processes run by that PASeq if you want to do so.
I decided to go with bezier paths so it should be 'quite' reproducible. If you run the generate paths command on a source photo you get a billion paths so you need to reduce the information. I used the cut-out filter in photoshop but I imagine the vectorizer could do the same. Simply reducing the source into a number of flat shapes. I used a couple of those one with little detail one with more. I loaded the first one into SA and switched to bezier mode and ran the paths>generate paths>from source edges command. Then ran path>path layer commands>size prune to get rid of the annoying little glitchy paths that get generated. I also hand deleted others that didn't look 'right'. Then ran path>path layer commands>simplify, this loosens the outline and makes it more sketchy. At this point if SA had better bezier tools you could add your own bezier lines and/or move and manipulate the lines already there - I didn't do that. I ran the path>paint path>full layer command with a 'pencil' preset [a pen looked too 'bald'.] Then I deleted the bezier layer and loaded my second cut-out image which had more detail and repeated the steps to get a second line drawing on top of the first. I then loaded my first cut out image the one with less detail and bigger colour blocks into source and use a graphic ink pen preset set to local image range in the paint fill apply>nib masking tab of the paint editor and loosely sketched some colour areas. *Hint* it didn't look a whole lot like Lincoln till I had done that!
At that point I liked it but decided to enhance and make it [to me] more interesting, so I added a paper texture, found a nice quote and finished up. Hope that inspires some more different approaches. [And a bezier re-vamp!] Enjoy the challenge - I'm off to see if I can refine this approach.
Interesting and different approach then what has been discussed before, and an interesting end result as well. Including an appropriate hand drawn quote too. Very well done.
Of course your use of Photoshop in what was essentially a pre-processing step for modifying a source image into something different that was then used as a source to make automatic bezier paths from totally disqualifies your result as a full solution of the challenge.
A successful solution only uses Studio Artist for all of the different components of the solution.
Would you mind trying that part again only using Studio Artist?
And then describing how you did it in Studio Artist.
You could use the vectorizer as you mentioned, or something like Color Simplify ip op. Other options might include Color Quantize or Watershed Regionize ip ops.
Simplify or Smooth ip ops can be used to significantly smooth a source image while maintaining it's edge structure, but they don't generate flat color region output.
And maybe you need to use manual selection afterwards to get rid of some of the flat colored regions generated by whatever color-flattening approach(s) you use?
Generating a raster flat color regions output (as opposed to an anti-aliased vector output) from either the Vectorizer or Color Simplify is going to make the selection process for getting rid of some flat colored regions in the result much easier. And probably makes more sense for your intended use, which is to have some flat colored regions that have very hard edges to use as input into the automatic bezier path generation algorithms available in the Path menu of Studio Artist.
I'm happy to provide other hints or tips on color flattening techniques in Studio Artist if you (or anyone else) needs them.
Thanks for that John,
This is just an interesting challenge. I am teetotal, so free beer is not a motivator : )
I used photshop because it is far simpler and quicker for what I had in mind- only a fool or someone out to prove something uses only a screwdriver to build a fence. SA is indispensable in my workflow but I would never use it exclusively, some features are just too clunky and awkward.
I posted merely to see if anyone else could take the bezier route any further. Every little helps right?
Let me make very clear that officially this is still just an 'intellectual challenge' only. As i mentioned previously, in the short term future, we hope to change that into an 'official contest' with 'official rules' and 'official prizes'. When that does happen, local micro brew beer will not be one of those official prizes. Sorry for any confusion about that.
I totally agree with you that when working, if you are familiar with tools for given tasks, it makes sense to just use them.
That said, the whole point of the Studio Artist intellectual challenge is really about us all learning more about Studio Artist features we maybe unfamiliar with, in addition to learning more about the analysis and interpretation of components of artistic style, and how to then recreate that using Studio Artist features.
That's why we want people to try to only use Studio Artist to pull off the entire challenge from start to finish.
My personal feeling is that generating flat colored regions from a source image is particularly easy to do in Studio Artist in many different ways.
So part of my goals for a challenge like this would be to get someone like yourself to learn enough about new Studio Artist features to also consider this part of the process an appropriate and easy to use tool for the job.
Or if you feel that is not the case, to understand why, and to work towards fixing that in the future.
Hope that made sense.
And i agree that your approach to generating the bezier paths automatically is a very insightful one. So thanks for sharing that with everyone.
At first glance - I would be tempted to say - this would be impossible to accomplish thru an automated process. But that thought is based on my lack of knowledge of what SA can do with automated processes.
The artistic, "organic" nature of the sample caricature - really makes automation a super challenge.
What follows will probably read a lot like a caricature "how to" primer along with my fiddling around in SA. I will try to associate each "artistic" step with a Studio Artist tool or feature - so that the how to is associated with a "with what" (tool). The process I used was entirely a hand applied one. Before i try any automated steps - I needed to see wether this look could be accomplished by hand.
I am attaching a zip file "CD_challenge2014" with the Paint Synth Presets I used in my attempts.
Note:
This sort of caricature would be an artists challenge - a major challenge. That is - it goes straight to the artists value as an artist and effective communicator - pulling from the real or photographed, then abstracting and codifying what seems important thru a completely different visual medium. Effective caricature (usually) requires a strong grounding in traditional drawing techniques (such as knowing how to indicate mass thru position, shading and overlap and knowing how to draw interlocking contours) and knowledge of the way the subject is built (anatomy) - and THEN - an ability to SEE and pick out significant features. General forms and telling details - and draw those in a lively way. The SEEing may not be something that will work as an automation… Picking those features that makes something unique requires a lot of "processing".
There are some "no experience required" techniques that could be applied to making a caricature. Kimon Nicolaides made contour drawing without looking much at the drawing surface a popular drawing/seeing exercise… I suppose a raw caricature could come from very little practice using a technique like this. Potentially using tools in SA could result in some interesting variations...
I made two passes at a caricature.
One I just played around in SA and did not record. Sadly - this play pass had the better outcome (better caricature)
Here is what I would be doing to address this caricature assignment - using SA as my digital tool exclusively:
Examining Johns chosen style/source:
There are several dominating features in the sample image to take into account.
• The white background. Clean except for the marks made by the artist.
• The forms of the caricature are built up with a contour outline.
(There are a few hatch marks to bring some depth shading into the image - same as the color fills)
• The scanty watercolor fills are strategically placed accents.
The clean white backdrop and deliberately applied color fills are a serious detour into prepping a file.
A one click automation that would apply something that isn't already in a source file seems unlikely.
Thats what makes it a serious challenge ; )
Starting with a photographic source (the Lincoln image posted) - I would need to take several preparatory steps.
Cleaning up a source image:
The standard simplification and exaggeration (caricaturing) of salient features would need to be attempted. This would usually be done "mentally" - that is the artist would zero in on features that made the subject uniques and ignore details that might distract from finding that uniqueness.
To recreate the stripped down idea in SA:
The source would need to be as separated as possible from a background.
Simplifying features:
Taking away unimportant features is a must. like removing details of skin. not applying too much shading etc. This is where that white space (removing background and details) becomes so necessary.
Using SA tools:
I went to the SA Adjust tools and Luminance Adjust and Smart BW Threshold to get a line art like outline.
Exaggeration of features:
This can be accomplished with some warping tools [or even a hand applied smudge]. Tho a specific warp used on one face very probably won't work from one subject to the next.
Exaggeration might also come in the form of adding extra emphasis to character features. More wrinkles or wilder hair… This might be accomplished thru cloning of some things/areas… But also starts to suggest that to achieve all the merits of this look, automation might not be the ideal solution.
Using SA tools:
I used a lot of Warp/ Warp Sphere and Translate Local to tweak the image into something more caricatured.
The attached history (zipped) was what I did to my second attempt.
The tools used in the sample image look like a very fine crow quill or radiograph like pen (black lines) and water color.
The surface looks like moderately rough watercolor paper.
I would make my final digital surface size at least 3000 pixels on its longest side - at 72 dpi. If this was for print - as much as 4000 pixels and nothing lower than 150 dpi.
For hand drawing. I would use a Paint Synth Preset that mimicked the pen tool like one of the cd_ink_ballpoint (included) options. These are tweaks of Presets that were actually made some time in the past to look a bit like a crow quill pen.
(One note on these Presets. These were designed to be used to hand draw lines - there is some built in width variation. There is also a significant difference between drawing quickly and slowly with these. The slower the line is drawn the more raggedy. This might be altered with a change in the way the Preset behaves related to velocity… This may also be an area where another control that smooths a line post drawn would be very useful. I encountered some frustrating limitations with the way the presets i used worked the more I drew… Eventually the smooth liquid line I expected from the Presets began to stutter and straighten out. This performance degradation affected the outcome of the art generated.)
The line work of Johns sample, especially around the clothes is clearly meant to suggest volumes rather than be an imitation of actual clothes/features. This suggestion of forms would be very difficult to pull off thru automation - though a Warp of some kind could be used to push things around in a source to look like loosely applied lines…
The color fills could be easily hand applied with Paint Synth Presets like:
cd_grainball…, cd_watercolor…, cd_wet… (included with this post)
I did make a first pass without recording and steps - just experimenting.
Sadly - that first experiment was actually closer to what i would have wanted for the second pass that i did record.
Here are some samples:
This is a rough framework after doing some Adjust and Warp
I had to draw back in some details.
Compare that with the second pass Adjust and Warp:
My second pass retain more of the source - but has lost some liveliness and personality (if not Lincolns)
Here is a line pass on the first attempt:
The tools behaved for this pass.
Line art for the second pass:
The line art here is showing the performance problems. Straight lines especially and shaky lines.
I big factor in feeling the line when caricaturing is not worrying about how loose a line is being laid down. The stiffness and palsied feel here became an issue. I tightened up trying to control the stylus.
Color experiment:
Color pass on second attempt:
My final results for the second pass were not entirely satisfying.
The tweaks to make the photo look more like a caricature resulted in of all things - a very awkward representation.
I would normally make many attempts at finding a look for the final caricature… learning as I go. A first and second pass like the ones done in SA would be the weakest passes in a non digital process.
Along with not loosening up with these passes, I encountered some significant degradation of performance to the Presets while hand drawing (and undoing) the second pass that led to some serious stiffening of the line work. Not sure what caused this… Perhaps file size or many undos of recording and playing histories??
None of the watercolor like fills are quite what i would want either - just looking for a satisfying watercolor fill resulted in hours spent building on old presets and experimenting with new effects.
Water color in SA being a mania for me. Getting tools that look good in actual use is something I continue to pursue.
The "play" version clearly has more personality. Ok, for a cold pass with many tools I seldom use.
---
The way I built the files was very simple.
I started at the base layer with the Lincoln jpeg on the canvas.
I made the Adjustment passes on that image.
I then set that image as the source image and probably scaled the canvas up to 3000 pixels.
I added a Layer and set it to Alpha Full Off. This was my color fill layer
I added another layer same as above and used that for my line art.
I painted lines and fills by hand.
I strongly feel that SA can be a great artistic tool. An application an artist can jump right into and make art that could go out to traditional venues, Web and Print as well and film. Recreating traditional media may not be the very best use of the possibilities SA has to offer - but the fact that I can accomplish things that can live very closely with traditional media is impressive (great app) and empowering - a great digital tool I can use to illustrate in dozens of "styles".
let me know your thoughts and if any of the routines I used might spark some ideas or tool development.
I will see if I can post some further experiments.
adjusted01_history.zip
CD_challenge2014.zip
-
1
-
2
-
3
of 3 Next