Negative Prompts, Explained (With Examples That Work)

What a negative prompt does

A negative prompt is a second prompt that tells the model what not to include. During sampling, the model pushes the image away from those tokens with the same strength it pushes toward the positive prompt.

Result: fewer extra fingers, no random text, less oversaturation, better anatomy.

The all-purpose starter

Copy this into every generation until you know what you're doing:

blurry, distorted, low quality, bad anatomy, deformed, extra limbs, watermark, signature, text, jpeg artifacts, oversaturated, low resolution

That single line will lift the quality floor of nearly every image you generate.

Negative prompts by use case

Portraits: asymmetric eyes, crossed eyes, missing teeth, extra fingers, mangled hands, plastic skin

Product photography: text, logo, brand name, packaging copy, watermark, glare, reflection artifacts

Landscapes: people, cars, buildings, signage, low contrast, muddy shadows

Illustration: 3d render, photo, photorealistic, jpeg artifacts, muddy colors

Traps

• Too many words — after ~40 tokens the model starts ignoring them. Focus on 5–15 that matter.
• Contradicting your positive prompt — if your positive says "detailed skin," don't put "smooth skin" in the negative.
• Style words — negative prompts are for defects, not styles. Steer style in the positive.

Imagoat automates this

Imagoat's Prompt Builder generates a sensible negative prompt for you automatically based on the subject and style you chose. You can still edit it before you generate.