LESSON 17 · VALUE & LIGHT
Value Composition & Thumbnails
A picture has to work as an arrangement of three value masses before it can work as anything else. This lesson teaches you to read and design that arrangement, and then hands you the skill almost nobody teaches explicitly: thumbnail iteration. You are going to make twenty bad thumbnails, and that is the entire point.
50 min
TIME
Intermediate
LEVEL
Photoshop or ArtRage
SOFTWARE
What you'll learn
Reduce any image to three value masses and judge whether it reads
Give one value dominance instead of splitting the canvas evenly
Identify high-key, low-key and mid-key value distributions
Produce twenty thumbnails in a session without polishing any of them
Iterate deliberately by changing one variable at a time
Select the two or three thumbnails worth developing
Before you start
Finish first: L14, L16
- The thumbnail grid download, or a blank 2400x1800 canvas
- A timer
- One subject you actually want to draw - a scene, a character in a place, anything with a story in it
WATCH
Watch this first
Thumbnail Iteration
Ctrl+Paint — The core lesson: quantity, speed, and the willingness to throw work away. This is the video the whole lesson is built on.
Value Distribution
Ctrl+Paint — Explains why an even split of light and dark is the weakest possible arrangement.
How to Learn Value Composition EASY
Proko — A structured way to practise composition using existing artwork as your source material.
THE LESSON
Three values, not five
Lesson 14 gave you five values for rendering a form. Composition works coarser than that: three. Light, middle, dark. If you cannot describe your picture in three masses - where the light is, where the dark is, and what shape the middle makes - the composition has not been designed, it has just happened.
This reduction has a name in Japanese design, notan, and the test is simple. Shrink your image until it is a couple of centimetres across, or squint hard at it, and see whether it resolves into a clear arrangement of a few big shapes. If it turns into an even grey field, the picture has no composition, regardless of how well individual objects are drawn.
The most common failure is even distribution. Roughly equal amounts of light, middle and dark spread evenly across the canvas gives no hierarchy - the eye has nowhere to go, so it goes nowhere. Give one value clear dominance, let a second support it, and let the third appear only as a small accent. A mostly dark canvas with one brilliant light shape has a focal point automatically. A canvas that is one third each does not.
Value key
The overall range you choose is the value key. A high-key picture lives in the light half of the scale - misty, bright, gentle. A low-key picture lives in the dark half - night, storm, threat. A mid-key picture stays in the middle and relies on small accents at either end.
Choosing a key is a mood decision made before you draw, and it constrains everything afterwards. A low-key painting does not mean everything is black; it means your light family sits around value 3, and the one small area of value 1 you allow yourself becomes enormously powerful. Restriction creates impact - the Lesson 14 principle of reserving pure white and black, applied to the whole picture.
Where the focal point goes is then a value question rather than a detail question. Put your greatest value contrast at the place you want the viewer to look, add your hardest edges there as well from Lesson 16, and the eye will go there whether the viewer intends it or not.
Make twenty bad ones
Here is the part that is rarely taught out loud. The skill in thumbnailing is not draughtsmanship, it is iteration - producing a large number of quick, disposable versions and keeping the best. Most people make one thumbnail, decide it is fine, and spend eight hours rendering their first idea.
Your first idea is almost never your best. It is your most available idea, which usually means the one you have seen most often - the composition everyone else thought of first too. Ideas four through twelve are where the interesting arrangements live, and you only reach them by passing through the boring ones. So the target is twenty, and most are supposed to be bad. Bad thumbnails are not failed work; they are the cost of arriving at a good one, at two minutes each.
Rules that make this work. Small: no bigger than 200 by 150 pixels. Fast: two minutes each, timed, and you stop when the timer stops. Three values only, straight from your swatches. Big brush, so detail is impossible. And the hard rule - never zoom in. Zooming in is how a thumbnail becomes a two-hour painting.
Do not judge as you go. Judging while generating kills the generating. Make all twenty first, and only then look at them.
Iterating on purpose
After the first five, stop making them up from nothing and start changing one variable at a time. Take a thumbnail that half works and produce four versions of it, each altering exactly one thing.
Useful variables, in rough order of power: change which value dominates, letting the dark mass take the frame instead of the light one. Change the crop - move in until the subject is cut by the frame edge, or pull out until it is small in a large space. Move the horizon. Flip the light so it comes from behind. Change the key from mid to low. Invert the notan: light subject on dark ground becomes dark on light.
Digitally this is cheap in a way it never was on paper. Duplicate the thumbnail layer with Ctrl+J or Cmd+J, move it to the next cell, and modify it rather than redrawing. Ctrl+I or Cmd+I inverts a layer, giving you the reversed notan in one keystroke, and Free Transform tests a different crop in seconds.
When you have twenty, zoom the whole sheet out until each thumbnail is about the size of a postage stamp, or step back a metre from the display. The two or three that still read clearly at that size are your candidates. Nothing else matters at this stage - not neatness, not draughtsmanship, not how much you liked drawing it. Only whether it reads.
What this sets up
This is the last lesson in the Value and Light module, and everything in it points forward. A composition designed in three values gives you the map, the five-value system fills it in, form and cast shadow give each object solidity, and edge control tells the viewer where to look.
When Module 5 introduces colour, none of this gets replaced - colour sits on top of a value structure that is already sound. Keep your best thumbnail from today. You will paint it properly later.
Never zoom in
Set the whole thumbnail sheet to Fit on Screen and leave it there for the entire session. The instant you zoom into one cell you have stopped thumbnailing and started illustrating, and you will produce three thumbnails instead of twenty.
PRACTICE · 40 MIN
Twenty thumbnails, three values, forty minutes
- Open the twenty-cell grid, or set up a 2400x1800 canvas with twenty cells at 200x150 each.
- Load three swatches only: value 1, value 3 and value 5 from your L14 scale. No other values are permitted today.
- Pick one subject and write it in the corner of the canvas - for example, a figure walking away down a corridor.
- Thumbnails 1 to 5: invent freely. Two minutes each, timer running, brush large enough that no detail is possible.
- Thumbnails 6 to 15: pick two of the first five and iterate. Change exactly one variable per thumbnail - dominance, crop, horizon height, light direction, or value key. Note the changed variable under each cell.
- Thumbnails 16 to 20: use Ctrl+J to duplicate your best three so far, then invert one with Ctrl+I and re-crop the others with Free Transform.
- Do not zoom in at any point. Do not tidy anything.
- Set the sheet to Fit on Screen, stand back a metre, and mark the three that still read as clear shapes.
- Redraw only the single best one at 1200x900, still in three values only, taking ten minutes.
Done when: Twenty completed thumbnails on one sheet, three of them marked as readable at postage-stamp size, and one enlarged three-value composition saved for the colour module.
Practice files
Download these and work along with the lesson.
Twenty-Cell Thumbnail Grid
2400x1800 px PSD, 300 dpi. BACKGROUND #FFFFFF locked. Locked layer GRID drawing twenty 200x150 px cells in a 5x4 arrangement with 40 px gutters, 1 px #BBBBBB borders, and a small numbered label 1 to 20 above each cell plus a blank 200x24 px caption bar below each for noting the variable changed. Twenty empty layers named THUMB 01 to THUMB 20. A locked layer SUBJECT with the placeholder text 'Subject:' in 40 px sans-serif in the top margin. Three-value swatch group embedded containing #E8E8E8, #808080 and #1A1A1A.
Notan and Value Distribution Plates
1920x1080 PNG, #F2F0EC background. Top row: four compositions of the same scene reduced to three flat values, showing light dominant, dark dominant, mid dominant, and an even split, with the even split marked 'weakest'. Middle row: the same scene shown in high key, mid key and low key, each with its value range indicated by a small bar chart against the five-step scale. Bottom row: eight thumbnails at 200x150 px demonstrating one variable changed per step - crop, horizon, light direction, inverted notan - with the changed variable captioned in 20 px type.
CHECK YOURSELF
Can you answer these?
Answer out loud first, then open each one to compare.
Can you describe a finished painting in three value masses and say which one dominates?
The check is that you can look at a finished painting, reduce it to light, middle and dark, and say which of the three owns most of the canvas. If you cannot describe it that way, the composition was not designed, it just happened. Shrink the image to a couple of centimetres across or squint hard at it — that is the notan test — and see whether it resolves into a clear arrangement of a few big shapes. If it turns into an even grey field, there is no composition there, however well the individual objects are drawn.
Can you explain why an even split of light, mid and dark is the weakest arrangement?
Because equal amounts of light, mid and dark spread evenly give the eye no hierarchy, so it has nowhere to go and goes nowhere. Composition needs one value clearly dominant, a second supporting it, and the third appearing only as a small accent — a mostly dark canvas with one brilliant light shape has a focal point automatically, and a canvas that is one third each does not. It is the Lesson 14 idea of reserving pure white and black, scaled up to the whole picture: restriction is what creates impact.
Can you produce twenty thumbnails in forty minutes without zooming in once?
This is a discipline check more than a drawing one. The rules that make it possible are: no bigger than 200 by 150 pixels, two minutes each with a timer, three values only straight from your swatches, and a brush too big for detail. The hard one is never zooming in — set the sheet to Fit on Screen and leave it there, because the instant you zoom into one cell you have started illustrating and you will finish with three thumbnails instead of twenty. Do not judge as you go either; judging while generating kills the generating, so make all twenty first and only then look at them.
Can you name five variables you can change to iterate on a composition you already have?
Which value dominates, the crop, the horizon height, the light direction, and the value key — with inverting the notan as a sixth. The point of naming them is that after the first five thumbnails you stop inventing from nothing and start changing exactly one variable at a time, which is what turns twenty thumbnails into a real search rather than twenty unrelated guesses. Digitally each of these is cheap: Ctrl+J (Cmd+J) duplicates a thumbnail into the next cell to modify, Ctrl+I (Cmd+I) inverts a layer for the reversed notan in one keystroke, and Free Transform tests a different crop in seconds.
Go deeper
Ctrl+Paint library
https://www.ctrlpaint.com/library
Winged Canvas digital art resources
https://www.wingedcanvas.com/digital-art-resources
Drawspace
https://www.drawspace.com/
← Previous
Rendering Form
Next →
Hue, Saturation, Value