Stop Naming Artists
The fastest way to get a forgettable image is to type "in the style of [famous artist]." It's lazy, it's legally murky in many jurisdictions, and worst of all, it produces the exact same image everyone else is making. Tools have been trained on millions of "in the style of Van Gogh" prompts, so the outputs have converged into a beige average of what the model thinks that name means.
There's a better move: describe the qualities you actually want. If you wanted Van Gogh, you wanted thick impasto brushstrokes, swirling directional texture, high-chroma blues and yellows, and a slightly elevated horizon line. Now you have a prompt you can tweak, mix, and own. You're directing instead of name-dropping.
This chapter is about building that vocabulary so you can describe any aesthetic from scratch.
The Four Levers of Style
Every visual style decomposes into four levers. Pull them independently and you can describe almost anything.
Medium
What is this image pretending to be made of? Oil paint, watercolor, charcoal, woodblock print, gouache, screen print, 35mm film, polaroid, 3D render, claymation still, vector illustration, pencil sketch, ink wash. Medium is the single most powerful word in your prompt because it constrains texture, edge quality, and color behavior all at once.
A red fox in a snowy forest, charcoal and chalk on toned grey paper,
visible smudges and fingerprints
Swap "charcoal and chalk" for "thin watercolor washes with wet edges" and you've changed everything without touching the subject.
Era
Time period is shorthand for a thousand technical decisions. "1970s magazine ad" implies grainy film, slightly warm color cast, Helvetica-adjacent type if any, a specific kind of photo composition. "1920s Bauhaus poster" implies flat color, geometric shapes, sans-serif type, red and black. "Y2K web graphic" implies chrome gradients, lens flares, glossy plastic.
Use eras as compression. Instead of listing twelve technical attributes, say "shot on expired 1990s color film" and the model fills in grain, color shift, and contrast for you.
Palette
Don't say "colorful." Say which colors, in what proportion, with what relationship. A few useful framings:
- Limited palette: "two-color palette of deep teal and burnt orange"
- Temperature: "cool palette dominated by blues and slate greys, single warm accent"
- Color theory: "analogous palette of yellow, yellow-green, green" or "split-complementary with a red-orange subject against teal and blue-green background"
- Mood-by-hue: "muted dusty pastels, low saturation, high value"
Portrait of a violinist mid-performance, three-color palette: deep
maroon, ivory, and brass gold, dramatic single light source, soft
shadows, oil painting on linen
Mood
Mood is what the image feels like before you can describe what's in it. "Melancholic," "uneasy," "triumphant," "intimate," "lonely," "playful," "sacred," "clinical." Pair mood words with how they're achieved: "uneasy: low-contrast, off-center subject, heavy negative space above." This forces you to articulate the mechanism, not just the vibe.
Building Your Vocabulary
Generic words produce generic images. "Beautiful" and "stunning" are noise — the model has no shared definition of them. Specific words produce specific images.
Keep a running document — call it your style sheet — where you collect words that worked. When you generate something that nails the look you wanted, write down the exact phrase that made the difference. Over a few weeks you'll build a personal dictionary of aesthetics you can recombine on demand.
Some categories worth populating:
- Light: "rim light," "raking light," "overcast soft light," "single candle source," "blown-out window light"
- Texture: "halftone dots," "linen weave visible," "fine film grain," "rough watercolor paper"
- Composition cues: "tight crop," "wide negative space," "shallow depth of field"
- Color descriptors: "muted," "high-chroma," "desaturated except for one accent," "bleached"
If you take one course alongside this chapter, AI Writing & Content Creation is useful — the same precision that makes a good written prompt makes a good visual one.
Using Reference Images
Most major tools now accept image inputs as style or composition references. This is where you graduate from describing to showing.
Two ways to use them:
-
Style reference: feed in a moodboard image — a vintage book cover, a frame from a film, a photo you took — and instruct the tool to borrow its color, texture, and lighting, not its subject. In Midjourney this is
--sref, in Flux and SDXL it's IP-Adapter or style LoRAs, in ChatGPT/DALL·E you upload directly. -
Composition reference: a rough sketch or an existing photo to lock down the layout while you change the style and subject.
Build your own moodboards. Photos you take with your phone of paint chips, fabric, lichen on a rock, magazine spreads in a thrift store — these are stronger references than anything you'll find on Pinterest, because they're already filtered through your eye.
Avoid scraping individual living artists' work as references for commercial output. Use found imagery, public domain archives (the New York Public Library, the Met, Europeana), your own photos, and your own sketches.
A Worked Example
Say you want a book cover for a story set in a coastal Greek village at dusk. Bad prompt: "beautiful Greek village at sunset, dramatic, cinematic." Reframed through the four levers:
Coastal Greek village at dusk, whitewashed buildings on a hillside,
gouache illustration with visible brushwork, palette of dusty cobalt,
warm terracotta, and pale bone, low-saturation evening light,
single warm window glow as accent, 1960s travel poster sensibility,
slight paper grain, calm and quietly nostalgic mood
That prompt makes one defensible decision after another — medium, palette, light, era, mood — none of them riding on someone else's name. You can now change "1960s travel poster" to "1990s indie graphic novel cover" and watch the entire piece transform while keeping the bones intact.
That ability — to dial style independently of subject — is what separates someone playing with a toy from someone building a portfolio.