What Is a Negative Prompt?
Stable Diffusion makes images from words. You type what you want. The AI draws it.
But there is second box too. This box is for negative prompt.
A negative prompt is different. Here you type what you do NOT want. The AI will try to avoid these things.
I found this box very powerful. Many people ignore it. That is a mistake. Your images get much better when you use it.
Why Negative Prompts Matter
Stable Diffusion is not perfect. Sometimes it makes ugly hands. Sometimes faces look strange. Sometimes the picture is blurry.
You can fix many of these problems. Just put the bad thing in the negative box.
For example, if hands look bad, you write "bad hands" in negative prompt. The AI now tries to make better hands.
In my experience, this small step saves lot of time. Before, I did many tries to get good image. Now I get it faster.
Basic Negative Prompt Words
Here are words I use most of the time. These work for almost every image.
- blurry
- low quality
- bad anatomy
- extra fingers
- deformed
- ugly
- watermark
- text
You can copy these and use them together. Just separate with comma.
Example: blurry, low quality, bad anatomy, extra fingers, deformed, watermark
This is good starting point. It cleans up most problems.
How to Write a Negative Prompt Step by Step
Let me show you simple way.
First, write your normal prompt. Like "a woman standing in a garden, sunny day."
Second, look at the result. What is wrong? Maybe the hands are bad. Maybe the picture looks like cartoon but you want real photo.
Third, add the bad thing to negative box. If you want real photo, add "cartoon, anime, drawing, painting" to negative.
Fourth, run again. The image will change.
I do this many times. Each time I fix one problem. This way I learn what each word does.
Negative Prompts for Portraits
Portraits are hard. Faces need to look right.
When I make portraits, I use these negative words:
bad eyes, cross eye, deformed face, extra limbs, mutated hands, poorly drawn face, disfigured, long neck
These help a lot. Eyes are the biggest problem in AI faces. This list fixes most of them.
Also, if the skin looks too smooth and fake, add "plastic skin, airbrushed." Then the skin looks more real.
Negative Prompts for Landscapes
Landscapes are easier. But they still have problems.
Sometimes there are strange buildings. Sometimes the sky looks flat.
For nature scenes, I use:
people, buildings, text, watermark, blurry, oversaturated, low detail
If you want empty nature with no people, "people" in negative is very helpful. It removes humans from the scene.
Do Not Use Too Many Words
This is a mistake I made when I start. I put fifty words in negative box. I thought more is better.
But it is not. Too many words confuse the AI. The image can get worse.
My advice: start small. Use five or ten words. Add more only when you need.
Watch what happens with each change. This way you understand your prompt.
Use Weights for Strong Control
Stable Diffusion lets you make words stronger. You use brackets and numbers.
Like this: (blurry:1.4)
The number 1.4 makes the word stronger. So the AI works harder to avoid blurry.
But be careful. If number is too high, like 2.0, it can break the image. I keep numbers between 1.1 and 1.5. This is safe zone.
My Simple Negative Prompt Template
Here is template I use almost every time. You can copy it.
lowres, bad anatomy, bad hands, text, error, missing fingers, extra digit, fewer digits, cropped, worst quality, low quality, jpeg artifacts, signature, watermark, blurry
This is my base. Then I add extra words for the specific image. It works for me very well.
Final Tips
Negative prompts take practice. You will not get perfect result first time. That is normal.
Here is what I learn. Keep a note file. When you find good negative words, save them. Next time you reuse them.
Also, different models work different. A negative word that works in one model may not work in other. Test and see.
Start simple. Add slowly. Save what works. This is the whole secret.
Try it today. Your images will look much better. I promise.