What is a negative prompt and what is it good for?

A negative prompt in Stable Diffusion is an input that describes which elements or characteristics a generated image should NOT include. This helps to avoid unwanted details. Here are some typical examples of negative prompts:

General Negative Prompts

• Unwanted Elements: low quality, blurry, watermark, text, logo
• Unnatural Features: distorted faces, extra limbs, missing fingers, unnatural anatomy
• Style or Mood: dark, gloomy, oversaturated, noisy, cartoonish (if realism is desired)
• Incorrect Details: wrong perspective, bad proportions, overexposure, artifacts

You can find suggestions for negative prompts here.

Negative Prompts for Specific Scenarios

1. Portraits: crooked eyes, asymmetrical features, distorted face, extra fingers, bad lighting
2. Landscapes: cloudy sky, unnatural colors, pixelated, unrealistic vegetation
3. Objects: rusty, broken, low-poly, unrealistic shadows, flat texture

Concrete Examples for Stable Diffusion

• Realistic Portraits: cartoonish, blurry, low-res, distorted, asymmetrical, extra fingers
• Landscape Scenes: overexposed, dark, unnatural colors, low detail, noisy
• Futuristic Architecture: old, traditional, medieval, broken, messy, unclean lines

A typical prompt set in Stable Diffusion might look like this:

Prompt Structure

• Positive Prompt: a beautiful woman with long hair, photorealistic, cinematic lighting, detailed eyes
• Negative Prompt: blurry, extra limbs, distorted face, low resolution, bad anatomy, cartoonish

Note: The negative prompt can often include as many details as necessary to minimize errors. Experimenting with specific terms can significantly influence results depending on the model.

Comentarii

Total 0 posts

Scrie un comentariu

Textul dumneavoastră:*


Vă rugăm să răspundeți la o întrebare (protecție împotriva spamului):*

Care dintre aceste animale trăiește în apă?

Câine
Vacă
Pește
Pasăre

* Introducere obligatorie