Negative Prompts Explained: How to Tell AI Image Tools What to Leave Out
What a negative prompt is, how it works in Midjourney and Stable Diffusion, and ready-to-use negative prompt lists for portraits, products, logos and AI video.
- Published
- Reading time
- 2 min
A negative prompt tells an image or video model what you don't want in the result: blurry details, extra fingers, text, watermarks, a cluttered background. It's the quickest fix for the same mistakes appearing again and again.
How negative prompts work
Image models build a picture step by step, nudging it toward your prompt. A negative prompt gives them a second direction: away from the things you list. That's why short lists of concrete things work best. "Ugly" is vague; "extra fingers, distorted hands" is something the model can steer away from.
Different tools handle it differently:
| Tool | How to add a negative prompt |
|---|---|
| Midjourney | Add --no at the end, followed by the items: --no text, watermark, people |
| Stable Diffusion (and most interfaces built on it) | A separate "negative prompt" field |
| Chat-based image tools (ChatGPT, Gemini) | No separate field; write it into the prompt: "Avoid text, logos and extra objects." |
| AI video tools | Many have a negative prompt field; otherwise add an "Avoid:" line at the end of the prompt |
Tool features change often, so check the current documentation of the tool you use.
Rules that make negative prompts work
- List objects and defects, not adjectives. "Watermark, signature, text" works better than "bad, low quality".
- Keep it short. Five to ten items is usually enough. Huge lists can make images flat and lifeless.
- Don't repeat the positive prompt in negative form. If you asked for "a white background", you don't also need
--no colored background. - Fix the main prompt first. If the subject is wrong, a negative prompt won't save it. Use negatives to remove recurring side effects.
- Change one thing at a time so you can see what actually helped.
Ready-to-use negative prompt lists
Portraits and people
extra fingers, distorted hands, extra limbs, crossed eyes, blurry face, plastic skin, heavy makeup, watermark, text, cropped head
Product photos
text, logo, watermark, extra objects, clutter, harsh shadows, reflections of the photographer, distorted label, fingerprints
For the full product photo setup, see our studio product shot prompt.
Logos and icons
gradient, shadow, 3d, photo, mockup, background texture, small details, multiple logos, text
It pairs well with the Midjourney logo design prompt.
Landscapes and interiors
people, cars, power lines, signs, text, watermark, oversaturated colors, fisheye distortion
AI video
subtitles, on-screen text, sudden cuts, camera shake, morphing objects, extra people, distorted hands, flickering
Our video prompts, like the cinematic product reveal, already end with an "Avoid" line built this way.
Troubleshooting
- The unwanted thing still appears. Make the main prompt more specific about what should be there instead. "A plain white wall behind the product" beats
--no clutter. - The image got dull. Your list is too long or too vague. Cut it to the three worst problems.
- Text keeps appearing. Remove any words from the positive prompt that sound like labels or signs (for example "poster" or "packaging with a label").
Browse more ready-made Midjourney prompts and AI video prompts in our library.
Frequently asked questions
What is a negative prompt?
A negative prompt tells an AI image or video model what you do not want in the result, such as text, watermarks, extra fingers or a cluttered background.
How do I use a negative prompt in Midjourney?
Add the --no parameter at the end of your prompt, followed by the things to exclude, for example: --no text, watermark, people.
Why is my negative prompt not working?
The list may be too vague or too long, or the main prompt may still suggest the unwanted thing. Make the positive prompt more specific about what should appear instead.