Image model · free · no sign-up

Midjourney
prompt generator.

Midjourney rewards short, concrete phrases over long sentences. The most important ideas should come first, and settings like aspect ratio or excluded elements belong in parameters at the end of the prompt, not in the description.

VisualPrompts turns your selections into exactly that shape: a comma-separated description ordered by importance, followed by --ar for the aspect ratio, --style raw for photographic realism and a --no list built from your negatives.

Build a Midjourney prompt Start from the example

Example outputGolden hour street portrait

lifestyle candid, waist-up medium shot, late 20s adult woman, walking toward camera, soft subtle smile, looking away from camera, messy textured hair, cream oversized knit sweater, city street, blurred bokeh city lights background, golden hour, golden hour sunlight, key light from behind the subject, romantic, dreamy, shot on full-frame DSLR, 85mm portrait lens, shallow depth of field, creamy bokeh, warm natural grading, photorealistic --ar 4:5 --style raw --v 7 --no CGI, 3D render, illustration, anime, plastic skin, over-smoothed face, text, captions, watermarks, logos, UI overlays, extra fingers

87 words

How to write better Midjourney prompts.

Lead with the subject and shot

Words near the start carry more weight. Put the genre, shot size and subject first, and details such as grain or palette near the end.

Keep it under ~60 words

Midjourney gives diminishing weight to long prompts. If the generator warns you about length, clear minor fields rather than adding more adjectives.

Use --no instead of “no …” in the text

Writing “no text” in the description can actually add text. Excluded elements belong in the --no parameter, which the generator fills from your negatives.

Add --style raw for photos

Raw mode reduces Midjourney’s default stylization, which helps photorealistic portraits and product shots look like real photographs.

References go in --oref or --sref

To keep a character or a look from an image, attach it with Omni Reference (--oref) or a style reference (--sref) instead of describing identity changes in words.