Introducing Synthography: A New Term for AI Art

by Benji Friedman

We are in the middle of a fundamental shift in how images are made. For the first time in history, a person can describe a scene, a feeling, or an impossible world — and a machine will render it into existence. This is not photography. It is not illustration. It is something new, and it needs its own name.

That name is Synthography.

Why a New Word?

The term “AI art” is broad to the point of meaninglessness. It could refer to a neural style transfer filter on a phone, a generative adversarial network producing faces, or a diffusion model conjuring landscapes from text. These are wildly different processes with different creative implications. We needed something more precise.

Elke Reinhuber proposed the term Synthography — a combination of “synthesis” and the Greek suffix “-graphy” (writing or recording). Where photography is “writing with light,” synthography is writing with synthesis. It describes the act of composing images through the orchestration of generative models — guiding an AI system to produce visual output through prompts, parameters, and iterative refinement.

The word captures something essential: this is a creative act, but the medium is algorithmic synthesis rather than pigment, film, or pixels placed by hand.

The Synthographer’s Craft

There is a misconception that generative AI art requires no skill — that you type a sentence and receive a masterpiece. Anyone who has spent serious time with these tools knows this is not true. The gap between a novice prompt and a refined synthographic work is enormous.

A synthographer develops intuition for how models interpret language, how different parameters shape output, and how to iterate toward a vision that may not have existed before the process began. It involves:

• Prompt architecture — constructing language that guides the model toward specific aesthetic territories
• Parameter exploration — manipulating CFG scale, sampling methods, seed values, and model weights to control the character of the output
• Curation and iteration — generating hundreds of outputs and selecting the ones that resonate, then refining further
• Post-processing — compositing, upscaling, color grading, and combining outputs into final works
• Model understanding — knowing the training data, biases, and tendencies of different models to work with (or against) them intentionally

What Makes Synthography Different

Synthography occupies a unique position in the history of image-making. It shares qualities with photography, printmaking, and digital art, but it is none of these things.

Like photography, it involves capturing something — but instead of capturing light reflected from physical objects, it captures patterns learned from millions of images, recombined into something that never existed. Like printmaking, each generation is a variation — the same prompt can yield wildly different results depending on the seed. And like digital art, the final output is pixels on a screen — but the process of arriving at those pixels is fundamentally different from placing them by hand.

What truly distinguishes synthography is its relationship to latent space— the vast, high-dimensional landscape of possible images encoded within a trained model. The synthographer navigates this space, exploring territories that no human eye has seen, finding images that exist as potentials within the model’s learned understanding of visual reality.

Low CFG: Letting the Model Speak

One of the most interesting frontiers in synthography is the exploration of low and zero CFG (Classifier-Free Guidance) images. CFG controls how closely the model adheres to a text prompt. At high values, the model tries hard to match your words. At low values, it drifts toward its own internal logic — producing images that feel like the model’s dreams.

These low-CFG outputs are proto-images: summations of visual concepts compressed from billions of training images, expressed without the constraint of human language. They reveal something about the nature of the model itself — its biases, its tendencies, its strange aesthetic preferences. This is synthography at its most exploratory, where the artist steps back and lets the medium assert itself.

Synthography as Exploration

Traditional art begins with a vision and works toward its realization. Synthography often works in reverse — you begin with a process and discover the vision within the output. The best synthographic work comes from artists who treat the model as a collaborator rather than a tool, who are willing to be surprised, and who develop the taste to recognize when something unexpected is also something meaningful.

My own practice spans series like Plant People, Forest Mazes, Lamassu, Flower Trash, and Low CFG explorations — each representing a different approach to navigating generative space. Some are tightly prompted, others are wild experiments in letting the model run free.

The Democratization Question

Synthography has lowered the barrier to image creation in a way that no previous technology has. Anyone with a computer can now produce compelling visual work. This is simultaneously exciting and unsettling — exciting because it means more people can participate in visual culture as creators, unsettling because it challenges long-held assumptions about what makes art valuable.

But accessibility has never diminished an art form. Photography faced the same criticism when it emerged — that it was too easy, too mechanical, too removed from “real” artistry. Today, photography is one of the most respected and expressive art forms in existence. Synthography will follow a similar arc. The tools will become ubiquitous, and the artists who push them furthest will define the medium.

Looking Forward

We are still in the earliest days of synthography. The models are improving rapidly — becoming more controllable, more expressive, more capable of realizing specific visions. New techniques emerge constantly: ControlNet, IP-Adapter, regional prompting, model merging, LoRA training. Each one expands the synthographer’s toolkit and opens new creative possibilities.

The term synthography gives us a way to talk about this work with precision. It acknowledges that this is a distinct creative practice with its own skills, aesthetics, and traditions — not a subset of photography, not a replacement for illustration, but something genuinely new in the long history of human image-making.

The images on this page represent my own explorations in this space. They are synthographs — made through the synthesis of language, algorithm, and artistic intention.

Explore more of my synthography work: AI Art · Plant People · Forest Mazes · Lamassu · Low CFG 2025 · Flower Trash