deut.li
Login
← Back

$27.2 Million for a Foot Sketch

Why Michelangelo was the first prompt engineer. A masterclass on Subject-to-Background Ratio and controlling attention through optics.

Neural Briefing

In early 2026, the art market experienced a shock comparable to a meteorite impact. At Christie's auction, a new record was set for Old Master graphics. A small, time-yellowed sheet of paper measuring just 13 by 14 centimeters went under the hammer for an incredible $27.2 million.

What is depicted on this sheet? A foot is drawn there. A right male foot. In red chalk.

The owner who discovered this drawing didn't even suspect its value. It was just an old sketch. Experts, however, confirmed: before us is a preparatory study by Michelangelo Buonarroti, created approximately in 1511–1512. Why is it rarest? Michelangelo destroyed sketches. He tried to hide his recipe for success from descendants. His searches. His findings, from which he later assembled masterpieces.

Why did he draw this?

"Of course, so his current owner could become a multimillionaire 500 years later."

Spoiler: not true.

Michelangelo didn't draw this foot for sale or for museums. He was solving a specific engineering-artistic task. This sketch is a preparation for one of the key figures on the Sistine Chapel ceiling, the Libyan Sibyl. On the fresco, the Sibyl is depicted in a complex turn (the famous figura serpentinata): she rises from her seat, holding a huge heavy book of prophecies in her hands, and simultaneously turns back.

All the physics of this movement, all the weight of the body and the book, all the kinetic energy of the complex pose falls on one fulcrum point — the toes of the right foot. The artist was looking for a way to tell this story through muscle tension. He needed not just to "stain the plaster" with paint. He needed to convince the viewer standing below of the reality of this book's weight. If he had erred in the anatomy of the foot, the entire monumental figure would have visually "crumbled".

Attention Management: A 500-Year-Old Technology

Michelangelo understood what suspiciously many content creators forget today. To control the viewer's attention, one must control ratios. One of the main tools artists have used for the last half-millennium is the Subject-to-Background Ratio. This is a mechanism that allows you to instantly place accents without saying a word.

What is "Subject-to-Background Ratio"?

Let's discard metaphysics. This is a purely geometric concept. It is literally the answer to the question: how much space by area (in percentages of pixels or canvas) does your Hero occupy in relation to the Environment.

Of course, in painting and photography, there are also color contrasts (red on green) and tonal relationships (light on dark). But we will leave them in parentheses for now. Geometry is primary.

  • If the object occupies 5% of the area — this is a story about the Environment, Loneliness, or Scale.
  • If the object occupies 90% of the area — this is a story about Detail, Emotion, or Texture.

Optics vs. The Pencil

How do professionals achieve this?

  • The Artist (Michelangelo): Simply draws the object larger or smaller. He "zooms" the image in his head.
  • The Photographer: Uses physics. The ratio is determined by the choice of lens (angle of view) and the physical distance from the camera to the object and from the object to the background. Want more background? Take a wide-angle lens (24mm) or step back. Want only the face? Take a telephoto lens (85mm+) or step closer.

The Problem of Generative Images

In 2026, when we create images in neural networks (Generative AI), the laws of composition have not changed. The physics of light and optics remained the same, even if simulated by an algorithm. A professional differs from a "dummy" in that they not only know about the subject/background ratio but also consciously work with it.

The dummy writes: "Beautiful girl on the street." The neural network gives an averaged result: a girl waist-deep, a bit of background, a bit of face. No accent. These images are not bad, because a human with their generalized angle of view of 46 degrees (or 50mm focal length) at a comfortable communication distance (without crossing personal space boundaries and without distancing) sees the object exactly like this. But the neural network didn't know what the generation customer actually wanted. What if they want to tell how wonderful this city is? Or how wonderful the girl is? Or what a wonderful earring hangs on her right ear?

To get such a result, you need to force the neural network to change optical settings. And here begins hell.

Syntactic Chaos: How it is done manually

In different neural networks, ratio control parameters are set completely differently. To get the same "close-up" effect (High Subject Ratio), you need to be a polyglot:

  1. Midjourney: Requires the technical language of a photographer. You need to write lens and distance parameters.
    • Prompt: --style raw --v 6.1 shot on 100mm macro lens, extreme close-up, depth of field
  2. DALL-E 3 / ChatGPT: Requires literary description. It often ignores technical terms, it needs to "explain" the scene.
    • Prompt: "A highly detailed macro shot showing only the texture of the skin."
  3. Stable Diffusion / Flux: Works on tags and weights.
    • Prompt: (extreme close-up:1.4), (macro photography:1.2), eyes focus

Neural networks might forgive a mistake in one word, but the more they conjecture, the further the result might turn out from the idea.

The deut.li Solution: Buttons Instead of Words

We created deut.li to save you from the need to remember these syntactic constructions. The interface has a simple set of buttons in the Framing block. You simply choose what you want to say.

  • Pressed the Tele button — the system formed the correct description in three dialects (Midjourney, Natural, Raw), emphasizing that you are interested in the object, and the background is secondary.
  • Pressed the Wide button — the system rewrote the prompt for "wide-angle optics" so that the background dominates the object.

The result is saved in the .deut format — a container file that stores the "DNA" of your image for the convenience of reuse or is copied to the clipboard for operative insertion into your generative model's interface.

Scenario: Changing the Accent Without Losing Essence

Imagine a situation: you generated the ideal character. But the art director wants two more images where this very character is of different sizes because the layout requires rhythm. The character is too far. The character is maximally close.

If you try to fix this in a normal chat ("make closer"), the neural network redraws everything. Yes, you use a reference image. Yes, you will try to persuade your neural network not to change the background, to play with scale. Sometimes you succeed, sometimes not. Read why this happens here (link to the first article in our blog).

In deut.li you do one action:

  1. Click the button in the Framing field: Wide.
  2. Click Generate.
  3. Send the prompt to your neural network. Result: Small character, background around.
  4. Click the button in the Framing field: Tele.
  5. Click Generate.
  6. Send the prompt to your neural network. Result: Character occupies almost the entire image field.

Since only a small detail responsible for optics changed in the prompt, while the other parameters (light, style, materials) remained frozen, the network renders the image maximally similar to the first variant. You get the same hero, the same atmosphere, but with the correct subject/background ratio.

Table: Optics as a Storyteller's Tool

We have prepared a table to help you understand how the choice of a "button" (focal length) changes the meaning of your image.

// OPTICS AS A STORYTELLER'S TOOL

FRAMING      | OPTICAL ANALOG  | SUBJECT/BG RATIO | EMOTIONAL EFFECT         | WHAT TO USE FOR
-------------|-----------------|------------------|--------------------------|----------------------------------
WIDE         | 14mm – 24mm     | Object < 10%     | Isolation / Epic Scale   | Hero lost in a vast world.
             |                 |                  |                          | Landscapes. Strong perspective.
-------------|-----------------|------------------|--------------------------|----------------------------------
STREET       | 35mm            | Object ~ 20-30%  | Context / Story          | Classic reportage. Hero +
             |                 |                  |                          | surroundings. Street scenes.
-------------|-----------------|------------------|--------------------------|----------------------------------
PORTRAIT     | 50mm            | Object ~ 50%     | Presence / Realism       | How the human eye sees.
             |                 |                  |                          | Lookbooks, calm scenes.
-------------|-----------------|------------------|--------------------------|----------------------------------
TELE         | 135mm – 200mm   | Object > 80%     | Compression / Drama      | Background compressed. Attention
             |                 |                  |                          | fixed on face. Fashion / portrait.
-------------|-----------------|------------------|--------------------------|----------------------------------
MACRO        | 100mm (Macro)   | Object = 100%    | Abstraction / Matter     | Only texture visible. Skin,
             |                 |                  |                          | chalk stroke, fabric fibers.

Michelangelo controlled the viewer's attention, which is why a piece of old paper with a sketch made in red chalk is worth such money 500 years later. The deut.li tool helps you control attention. Don't write essays. Press buttons.

Don't type. Snap it in.