August 14, 2005
Alias SketchBook Pro 2.0
I picked up Alias Sketchbook Pro 2.0 at the SIGGRAPH promotional price of $149, and I'm quite happy with it.
It's marketed for "sketching" but if you design your own pressure-sensitive brush and use the method I outline in the image above, you may find the software can achieve painterly blends.
Posted by digital artform at 11:30 PM | Comments (0)
July 26, 2005
Drawing from Memory
Draw something from reference. Let's say it's a skull.
Hide the reference and draw it again from memory. Focus on essential planes and basic forms, not little surface details.
You will be imediately confronted by what you don't know about it. Is the mastoid process above or below the teeth? What is directly above the last molars? Things like that.
Write your questions down.
When you use terms like "above" and "below", make sure your questions are phrased in terms of the imaginary axes of the object itself -- not the axes of the paper. If you ask questions like "In this particular view, is the base of the nose above the base of the back of the skull?" you will be memorizing a drawing -- not an object's form. The idea is not to memorize a series of drawings, but to learn the object itself, so that later you can use what you learned to draw it from any angle. In the photo above, the blue lines are not a fixed grid in paper-space; they are the edges of cubes. When you turn the skull off axis, mentally turn the cubes off axis as well.
Draw it again from reference. Answer your questions.
Repeat the process with and without reference until you have fewer and fewer questions and can draw the object well without reference.
Move on to another view of it and repeat the process until you don't need reference for any view at all.
You may be surprised at how fast you can develop a "conception of form" for a complex object.
The feedback question / answer loop is better than staring at an object and trying to memorize it through sheer force of will.
The nineteenth century French artist Horace Lecoq de Boisbaudran was a pioneering proponent of this general sort of learning method.
Posted by digital artform at 08:39 AM | Comments (2)
July 23, 2005
Lighting a Sphere
When rendering a sphere, people often put the highlight in the wrong place. The highlight does not belong "at noon."
It is true that a diffuse sphere, like the Moon, will be brightest where the sun shines most directly -- where it is "noon."
However, a sphere with any "shinyness" to it will have its hot spot not at "noon," but in the place on its surface midway between the angle to the light, and the angle to the viewer.
Posted by digital artform at 11:00 PM | Comments (2)
July 13, 2005
Vanishing Point Perspective Guide
You know your lines need to converge to a common vanishing point. But what do you do when that vanishing point is so far along the horizon that it is not within your composition?
Here's a guide I put together on how to plan a perspective grid in Photoshop when the vanishing points are far off the page.
First a little background on layers and reference points in Photoshop.
Make a straight line on an otherwise new, transparent layer. In this case I have made a somewhat diagonal line, so that the line's tips are in the upper left and lower right corners of the line's imaginary bounding box.
Hit ctrl-t (in Windows) to transform the layer containing the line. You'll see a bounding box around the line, and a special point in the middle of the box called the reference point. Hover your cursor near the edge of the box until to see a curved arrow. The curved arrow indicates that the layer can be rotated. Drag the layer. Notice how by default it rotates around the reference point in the center of the layer.
While you are transforming the layer, look up in the options box along the top of the Photoshop CS2 workspace. See that little tic-tac-toe board? That is one way of controling the location of a layer's reference point.
Hit the white square in the upper left corner of the tic-tac-toe board before you rotate your layer. See now how the line itself rotates about one of its own end points?
The line will rotate about the end point that is located in the upper left hand corner of its bounding box.
Can you guess where I'm going with this?
Enough groundwork. Let's begin:
Go to full screen mode.
Zoom way out and establish a distant vanishing point. Use your talent and instinct to figure where your vanishing point needs to be -- that is outside the scope of this tutorial.
You may find guides useful here. Guides are the blue lines you get when you drag off of the rulers at the edge of the image.
Establish a line from your vanishing point through your image and out the other side.
Hit ctrl-t (on Windows) to transform your line layer at will.
but make sure you use the reference point locator first -- you want to rotate the line about the layer's corner, not the layer's center!
Posted by digital artform at 12:29 PM | Comments (3)
June 14, 2005
Color Judgement in Context
For a really good example of why you should develop your painting in all areas at the same time and not just throw in a background later, check out illusion #3 of these three optical illusions from echalk.co.uk:
Even if you know it already, it will amaze you to see just how much context influences your color judgement.
...and on a somewhat related note, here's an old (pre-blog) page of mine on how to hone your color judgement using Photoshop and a digital camera.
Posted by digital artform at 08:49 AM | Comments (0)
June 10, 2005
Digital Color Mixing in Photoshop
When I paint in Photoshop, I blend colors using variable opacity under the control of a pressure-sensitive Wacom pen.
Grad-based blending from one color to another is not true digital color mixing, however. A linear blend from yellow to blue, for example, goes through middle gray without producing any sense of green.
If you want to simulate color mixing with opacity falloff blending from one color to another, you might want to help the blend along by adding a third color along the way.
Color mixing (which has the unfortunate and misleading name subtractive color mixing) is closer in behavior to multiplication, or the sandwiching of color filters. A yellow filter that only lets through pure yellow light (a single wavelength at around 570 nanometers) sandwiched with a blue (475 nm) filter will not create a green color, because neither filter allows green (510 nm) light to pass. It's the same deal with pigments.
A "sloppy" blue filter -- one that looks blue to the eye but lets other wavelengths slip through, sandwiched with a sloppy yellow filter, will produce a nice green, provided that there is plenty of green in each.
Real world pigments are like sloppy filters.
If real world paints could be manufactured with their reflective qualities confined to extremely narrow wavelengths, then they would look saturated and colorful but mix to produce black.
There is no exact mapping of RGB values to wavelengths. You could make conversion lookup tables based on the feelings of panels of human observers. For any given color as perceived by a person, there are many "wavelength recipes" that could be responsible for creating in that individual that particular color impression. How colors mix depends on what wavelengths are present in their wavelength recipe. When two colors mix, those wavelengths which both colors have in common survive the mix.
Multispectral sampling to represent colors with more than the usual 3 (red, green, and blue) is not a new idea in computer graphics. The ability to represent colors with arbitrarily many spectral samples (often 9 or 12 in practice) has been part of -- for example -- the Renderman Interface Specifications -- for some time now (see The Renderman Companion [1990] by Steve Upstill, page 42, for discussion of this topic). I'd like to think this idea has worked its way into Painter IX's color mixing by now, but I haven't seen any specific discussion to this efect.
If you'd like a good handle on color mixing, don't miss Blue and Yellow Don't Make Green by Michael Wilcox.
And pay a lengthy visit to these handprint media web pages on color and color theory.
Posted by digital artform at 03:12 PM | Comments (0)
June 05, 2005
Age Your Digital Paintings
A little crackle medium, available at most art and craft supply stores, can give your digital paintings an aged realism.
Lay out a layer of white acrylic paint onto a surface you don't mind ruining, and apply crackle medium according to the product instructions.
When the acrylic paint shrinks, it will crackle as if with age.
Use your favorite Photoshop blend modes to work the cracks into your digital paintings. For the skull example below, I didn't do too much more than just multiply, one image by the other. If you put more time and effort into combining your crackle paint image with your digital painting, you can achieve extremely realistic results.
You may even find the patterns useful in 3D CG for creating things like sun baked mud.
It's a handy texture to have around. I even used it in the banner at the top of this web site.
Posted by digital artform at 09:49 PM | Comments (3)
April 22, 2005
Painting with Photoshop
The two most important brushes in Photoshop are the hard-edged circle and the soft-edged circle.
The most important tool is a tablet with a pen.
Set your brush's opacity to respond to pen pressure.
Most things in the world have hard edges. You'll use your hard-edged brush a lot.
Many color transitions on or within objects are soft. You may find your soft-edged brush comes in handy in those cases -- but not as handy as you think -- don't over rely on it.
You can get a huge amount (maybe even most) of your work done with with a pressure-sensitive hard-edged circle.
How do you blend one color into another with a hard-edged brush?
Use a light touch (low pressure) and paint this color into that, and that color back into this.
Your two best hot keys are the size changers, ] for bigger size and [ for smaller size, and the eye dropper key (alt on Windows)
Don't eye-dropper colors from photos. Try to estimate them by eye. Feel free to sample your own colors from elsewhere in your painting.
Personally, I prefer not to let pressure control brush size; only opacity.
If you use the pressure sensitivity feature for any brush settings, take the time to go into the Wacom software and adjust the pressure sensitivity of the tip.
There is a slider that remaps pressure values through a curve. I like to set my pen tip to be one notch to the right of (firmer than) the default setting. I find this brings out the subtlety of the light touch, and helps prevent the tip from slamming from "barely on" to "fully on" without a smooth transition.
Using opacity falloff to model color mixing isn't "true" color mixing. The opacity blending method will only take you so far. If you feel there is a step missing in your blend, you can "help it out" by sticking the color into the blend yourself.
There are other ways to more accurately model digital color mixing within Photoshop. I talk more about that in my Digital Color Mixing with Photoshop post. Here's a sneak preview:
I don't actually use "multiply mixing" that much in Photoshop. It's not convenient. When I paint, I lay down direct color and opacity-blend it as described here. If I need to help a color mix along manually by introducing a third color, I do so.
If you don't like this style of painting and color mixing, look into Corel Painter as an alternative (and in many ways superior) digital painting application.
And don't miss Blue and Yellow Don't Make Green by Michael Wilcox.
Posted by digital artform at 10:26 AM | Comments (2)
January 15, 2005
Varied Line Weight in Illustrator
I was experimenting with techniques for creating an organic, varied line weight in vector apps like Adobe Illustrator and came up with this approach.
Cartoony illustration is not really my thing, but I thought it might be a good choice for a poster for local Jazz Festival. To get the varied line weight, I drew each line twice -- first forwards, then immediately backwards -- loosely retracing the first line. I let Illustrator fill the resulting closed loop. No stroke was used -- only fill. I was sort of going for a style like that in the old Anna Nicole Smith tv show main titles.
Posted by digital artform at 10:06 PM | Comments (0)
December 01, 2004
One Point Perspective
In one-point perspective, all parallel lines perpendicular to the image place recede to or radiate from a common point in the center of the image. That center can only be moved by cropping the final art in one place or another. The "speed" with which lines parallel to the image plane "pile up" can be altered by changing one's sense of camera lens and placement.
When talented artist Robert Chang mentioned on an art forum that he was experimenting with layouts to explore the depiction of depth in 1-point perspective for a new work of his, I decided to take a closer look at the issue myself.
Here is a typically dramatic one-point perspective layout. The figures are mapped with alpha-controlled transparency onto a vertical wall at one end of a long checkered corridor.
If I simply zoom in from a 20mm to a 40mm lens, the distortion will not change (why?), but the layout will enlarge. If I re-frame the figures by moving the camera further away, the layout seems to "flatten" a bit.
If I zoom in again, this time from a 40mm to an 80mm lens, the distortion again will not change (why?), and the layout again will simply enlarge. If I again re-frame the figures by moving the camera further away, the layout seems to "flatten" a bit more.
Notice how as the lens lengthens and the camera pulls away, the relative jump in depth from one checker square to the next diminishes. In all cases, however, the vanishing lines stay the same -- things just "stack up" on them differently.
Foreground figures courtesy of Robert Chang. Used with permission.
Posted by digital artform at 08:46 AM | Comments (0)
November 28, 2004
How to Draw Ellipses
I was intrigued by a tutorial I found on artist Scott Robertson's web site drawthrough.com on how to freehand draw ellpses. I found it extremely educational, and wanted to play around in 3D with what I learned there.
His first main assertion, which I see a lot of people get wrong, is that when a circle in 3D space projects into 2D as an ellipse, the center of the circle and the center of the ellipse are not the same point. That much I already knew.
His second assertion, which I found particularly interesting, is that in such a situation, the minor axis of the 2D ellipse is the same as (or is at least parallel to?) the normal axis of the original 3D circle. In the case of a car, the minor axis of the 2D ellipses is parallel to the axle of the car. In the case of an airplane, the minor axis of the propeller's ellipse in parallel to the drive shaft.
I like to test things. Here's what I got when I did.
My results agree pretty closely with his predictions. Close enough that I think I'll be mindful of whether or not I'm following his suggestions when I draw.
The results of my experiment didn't look "dead-on," however. My own inaccuracy orienting the red axis along a vertical line and in hand placing the 2D test ellipses might be one source of the error.
Another source might be in the way in which the original tutorial depicts objects in perspective. Even when Robertson draws the X- and Z-axes of an object in perspective, he maintains the Y-axis perfectly vertical, as if he were using a view camera, or a kind of draftsman's perspective not produced by typical cameras.
Posted by digital artform at 02:23 PM | Comments (0)
October 24, 2004
Avoiding Ugly Grads in 2D
If you paint a monochramatic gray underpainting and colorize it by simply multiplying it by a base color, you'll get a dull image. Even if you hand pick and paint every color in your image, if you rely only on the luminosity slider, you'll have the same problems. You see this luminosity slider reliance a lot, especially in the skin tones of the work of beginning painters.
When a form turns away from light it usually experiences a hue shift. This often happens simply because the fill light is a different color from the key light. Regardless of why it happens, however, it usually looks better when it happens. So when you paint or light 3D, identify or invent a reason to shift the hue warmer or cooler, and sell it.
Here's a related link:
Avoiding Ugly Color Falloff in 3D
Posted by digital artform at 11:12 AM | Comments (0)
October 16, 2004
Maya 3D Fake "NPR" for Print
I was curious to see to what extent 3D computer graphics could integrate with 2D digital work. I was especially keen to acieve styles in 3D that had a "Non-Photorealistic" (NPR) quality - more illustrative or painterly looking. I found I could get some interesting results by modelling 3D primitives in a fairly crude way, and then selectively blurring certain edges (such as the interior edges in these marshmallow fluff couds.)
I thought another promising direction might be to use Painter to fake together some rough "paintovers."
It's not "real" painting, and it's not real "NPR," either, but I see potential in the end results, especially for still images in a print context, and I may explore more along these lines in future.
Posted by digital artform at 09:23 PM | Comments (0)
October 13, 2004
Color vs Value in Painting
Artists stuggle over getting correct colors, yet often neglect to get correct values, in spite of the fact that of the two, values are more important than colors at revealing an object's form.
It's a good idea to break scenes up into 5 or so value levels, from black to white, and to try to determine with accuracy which portion of a painting falls into which value level.
Be careful when turning color images into black and white images using Photoshop. The software may not act as you expect. In particular, it will place the grayscale divisions in even steps, which may not suit your purposes. Don't rely on this technique in your painting practice; it's more a way of explaining an idea. Develop your own skills in determining what areas of your image deserve which values.
Some bright colors appear to be brighter to the human eye than other bright colors. Blues tend to look darkest. The common equation for converting RGB to Luminosity is the NTSC (National Television Standards Committee) formula:
luminosity = 0.299 x red + 0.587 x green + 0.114 x blue
If you go through the trouble of changing the image mode to grayscale, you will get what you expect you should get.
If you do not use the mode > grayscale technique, you will get surprising results.
If you merely desaturate the image, which is what most people do, you will get a mathematically plausible but psychologically wrong looking image in which all equally saturated colors become equal grays - regardless of their hues, or their effect on human perception.
Unexpected? As a painter, you will probably want to avoid this outcome.
UPDATE 5/26/05
Craig Mullins (screen name spooge demon, in this thread) is the master of this. Here's an excerpt from what Craig Mullins says on the matter of values:
"Decide what is in light and what is in shadow and don’t mix them up. Think like a comic artist. Two values, but if they are well thought out and designed and drawn they can look totally real. Think like that, but instead of making the light white and the shadow black, make the light a 7 and the shadow a 3. Then go ahead and use 5-10 in the light and 1-3 in the shadow to pull out sub forms. DO NOT use 1-5 in any part of the light, or use 5-10 in any areas of the dark. Keep you edges a little softer in the shadows, a little sharper in the light, you are done. (0 is black, 10 is white) Deciding what is in shadow and light for a particular object is pretty hard in words. I will leave that up to you and that is 99 percent of the struggle."
I try to learn from him.
UPDATE 12/10/05
I was re-reading Paul Haeberli's classic Grafica Obscura today and noticed that he makes the point that the usual NTSC luminosity weights are intended for gamma 2.2 work, not linear gamma 1.0 work. Here's an excerpt...
http://www.sgi.com/misc/grafica/matrix/index.html
Converting to Luminance
To convert a color image into a black and white image, this matrix is used:
float mat[4][4] = {
rwgt, rwgt, rwgt, 0.0,
gwgt, gwgt, gwgt, 0.0,
bwgt, bwgt, bwgt, 0.0,
0.0, 0.0, 0.0, 1.0,
};
Where rwgt is 0.3086, gwgt is 0.6094, and bwgt is 0.0820. This is the luminance vector. Notice here that we do not use the standard NTSC weights of 0.299, 0.587, and 0.114. The NTSC weights are only applicable to RGB colors in a gamma 2.2 color space. For linear RGB colors the values above are better.
In effect, this calculates:
tr = r*rwgt + g*gwgt + b*bwgt;
tg = r*rwgt + g*gwgt + b*bwgt;
tb = r*rwgt + g*gwgt + b*bwgt;
Posted by digital artform at 11:20 PM | Comments (2)