|
The prompt economy has spent the last three years selling the same idea. Longer is stronger. Pack in adjectives. Stack the camera specs. Add the lighting setup, the film stock, the artist references, the mood adjectives, the negative prompts, the magic phrases. Three hundred words per image. The longer the prompt, the more polished the output. That was the conventional wisdom and it has not held up.
I tested this carefully over the last two weeks. Forty matched pairs. The same intent expressed in a seven-word prompt and in a paragraph-length prompt, run through ChatGPT Images, Midjourney V8.1, and Z Image. The short prompts won thirty-two times out of forty. The long prompts won six times. Two were ties.
The pattern was consistent. Long prompts tend to confuse newer models because the models try to satisfy every clause, which produces composite images that feel overworked. Short prompts give the model room to make its own decisions inside a clear frame, which produces cleaner output. The trick is knowing which seven words to pick.
|
|
The seven words break into four structural slots. Subject, action or state, setting, and visual mode. Each slot does one job. Together they give the model exactly enough to make a strong decision and nothing more.
[subject] [action or state] [setting], [visual mode]
Subject is the noun the image is about. Action or state is what the subject is doing or how it exists. Setting is where the scene lives. Visual mode is the single most important word in the prompt, because it tells the model what kind of image this is. Photograph, painting, illustration, render, sketch. The visual mode controls everything else downstream.
A working seven-word prompt looks like this:
|
|
Each pair below is the same intent expressed in a short prompt and a long one. Run them yourself if you doubt the result. Run them in any image tool. The short ones consistently produce cleaner output.
PAIR 1 PORTRAIT
Short: Woman reading book sunlit window, photograph.
Long: A beautiful 30-year-old woman with brown hair sitting by a large window in a Brooklyn brownstone, holding an open hardcover book, soft natural light streaming in from the left, shot on a Canon R5 with 85mm f/1.4 lens, shallow depth of field, warm afternoon light, editorial photography style, film grain, the atmosphere of a quiet Sunday morning.
PAIR 2 PRODUCT
Short: Ceramic mug steaming wooden table, photograph.
Long: A handcrafted white ceramic mug with a matte finish sitting on a rustic wooden table, steam rising from freshly brewed coffee, morning sunlight casting soft shadows from the left, shot on medium format with 80mm lens at f/2.8, shallow depth of field, the warm minimalist aesthetic of a Scandinavian design magazine, slight film grain, editorial product photography.
PAIR 3 SCENE
Short: Empty cafe rainy afternoon, painterly illustration.
Long: An empty Parisian cafe interior on a rainy October afternoon, soft amber lighting from vintage Edison bulbs, wooden chairs and small marble tables, condensation on the windows, raindrops streaking the glass, melancholic atmosphere, painted in the style of Edward Hopper meets contemporary digital illustration, muted color palette of deep greens and warm browns, oil painting texture.
PAIR 4 SOCIAL POST
Short: Skincare bottle marble countertop, editorial photograph.
Long: A minimalist skincare bottle with frosted glass on a white Carrara marble countertop, soft diffused window light from above, eucalyptus leaves casting subtle shadows, the clean luxurious aesthetic of a high-end beauty brand Instagram feed, 1:1 square composition, negative space on the right, shot on Hasselblad with 80mm lens, modern minimalist beauty photography.
PAIR 5 ANIME
Short: Girl rooftop sunset city, anime illustration.
Long: A teenage girl with long black hair tied in a ponytail wearing a school uniform standing on the rooftop of a Tokyo apartment building during golden hour sunset, looking out over the city skyline, wind blowing her hair gently, warm orange and pink sky with scattered clouds, soft cel shading, shoujo anime style, detailed background with realistic urban architecture, melancholic mood.
|
|
Six of the forty pairs went to the long prompt. They had something in common. The intent was unusually specific. A recreation of a known photograph. A character with named features that had to be preserved. A complex multi-subject scene with required composition. In those cases, the extra clauses gave the model real information it needed.
The honest rule is that long prompts win when you can name what you want with precision. Short prompts win when you can describe it with restraint. Most of the time, restraint produces stronger output, because most prompts contain more wishes than the model can satisfy without compromising on each one.
The next time you write a prompt, try the seven-word version first. Run it. See what comes back. If it misses, add words. If it is close, refine the visual mode word at the end. Most images do not need the paragraph. They need the right seven words.
|
