Birth of Emoji: What Botticelli Taught Me

Birth of Emoji is built from 24,088 tiles drawing on 986 unique emoji. It required 31 artist regions — more than any Diamoji to date — and introduced full color calibration to the process for the first time.

Every Diamoji I've made has taught me something new. A Great Wave of Emoji was about structure and motion. A Starry Night of Emoji was about color and energy. Birth of Venus was about something harder to pin down: precision.

Birth of Emoji is complete. It is built from 24,088 tiles drawing on 986 unique emoji. It is the most technically complex Diamoji to date — the soft palette and dissolving edges of Botticelli's original pushed the process into new territory. This post is about what made it difficult and what I learned along the way.

The painting itself presented new challenges from the start.

Botticelli painted Venus with a soft, hazy quality that art historians call sfumato. Edges dissolve rather than define. The color palette is warm and subtle — creams, pinks, coral, and muted sage. There are no sharp transitions to anchor the system. Every prior Diamoji I made featured subject matter with strong structure: crashing waves, swirling night sky, a face with unmistakable geometry. Venus offered none of that.

The background added another layer of difficulty. A large portion of the canvas is open sky and water. These areas produce enormous tiles with relatively simple color. They carry enormous weight in the statistics I use to calibrate the piece. Getting the detail tiles right while those background tiles dominated every calculation took real work.

This project required more artist regions than any Diamoji to date.

Artist regions are how I direct the system's attention. I draw them by hand in Photoshop — bounding shapes around each figure, each area of the composition that needs to be handled differently. More regions means more control. It also means more decisions.

Birth of Venus has 31 artist regions. The figures alone account for most of them: Venus herself, the Hora figure on the right, Zephyr and Aura on the left, individual faces, hands, the shell at her feet, the flower-scattered shoreline. Each figure required its own region, sometimes more than one, to get the tile density right across the composition.

Artist regions designate where details are needed. The regions start as red rectangles and magenta line segments in Photoshop. The geometries are exported and fed into the rendering pipeline along with my additional settings about their reach and level of detail.

Botticelli's sfumato technique was the source of a specific technical problem. The system uses edge strength to decide how finely to subdivide each tile. Sfumato produces very soft edges by design — so the edge values across the figure were too low to trigger the fine subdivision I needed. I had to adjust how the system weighs edge signals specifically for this project to get the figure detail the painting deserves.

Birth of Venus was the first Diamoji to go through full color calibration.

Color in Diamojism is measured in LAB — a color space designed to match how human eyes perceive brightness and hue, rather than how screens produce them. After a major render, I analyze how closely the emoji layer matches the source painting across the entire LAB spectrum. When the numbers show a systematic bias — the output running too warm, too bright, too flat — I can compute a correction and apply it before the next render.

For Birth of Emoji, I needed one pass of calibration to bring the emoji palette into alignment with the painting's warm, skin-tone-heavy middle range. A second pass would have overcorrected. I could see it in the data: the numbers that had been slightly too high flipped negative, which meant the system had gone too far in the other direction.

Knowing when to stop is part of the craft.

The most important tuning parameter for Birth of Emoji after color calibration was a value that controls global color saturation. I tested it at four different settings. At 1.00, warm skin tones went muddy. At 1.02, the wind figures on the left read orange. At 1.01, the warmth held. That is the version you see in the finished work.