Stable Diffusion Negative Prompt for Human

When using AI image generation models like Stable Diffusion, it is important to provide both positive prompts describing what you want to see, as well as negative prompts specifying what you don’t want. Negative prompts give you more control over the output and help avoid generating strange or undesirable images.

In this article, I will provide useful negative prompt examples for Stable Diffusion focused on generating realistic and high-quality human images. Whether you are an artist, photographer, or just experimenting with AI for fun, these prompts will help guide the model.

Avoiding Unnatural Images

Here are some common negative prompts to avoid generating unnatural, mutated, or non-human images:

Deformed Body Parts

deformed, mutated, mutilated, distorted, disfigured 
extra limbs, missing limbs, floating limbs
disconnected body parts

Surreal Imagery

surreal, abstract, cubist, cartoon, cg, 3d, unreal
animate, non-human

Poor Image Quality

low quality, low resolution, pixelated, jpeg artifacts
blurry, unclear, out of focus, depth of field

Realistic Human Figures

When generating images of people, using negative prompts related to anatomy can help produce more realistic results:

Bad Anatomy

bad anatomy, wrong anatomy
bad proportions, disproportionate

Missing Body Parts

missing arms, missing legs, missing fingers
amputated, amputation

Extra Appendages

extra arms, third arm, extra legs  
too many fingers, fused fingers

Suitable Content

To avoid NSFW content, prohibited content, or themes you simply don’t want, add:

nsfw, nude, censored
gore, violence, horror
controversial, offensive

Conclusion

Negative prompts are a powerful way to guide AI image generation. Start by adding just a few general prompts and refine over time based on the results. The examples provided here for generating realistic human images are a good starting point.

Remember to also use positive prompts to indicate what you want to see. Balancing positive and negative prompts takes experimentation, but leads to better outputs.