How to fix these common issues in AI images
October 31, 2022
Seeing some funny issues in AI images? I will teach you how to fix some of the common ones.
We will use this Stable Diffusion GUI for this tutorial. See my quick start guide for setting up in Google’s cloud server.
It’s not unusual to see images with two heads connecting together in Stable Diffusion. It is usually caused by using portrait image size. To avoid it, use image size with 1:1 aspect ratio, for example 512x512.
Three ways to tackle the two-head issue:
- Generate multiple images and discard those with two heads.
- Reduce aspect ratio. The closer to 1:1 ratio, the less likely you would encounter the infamous two-head.
- Generate a full-body image. (See next section for techniques.)
Not showing full body
A common suggestion to get a full-body image is to use the keyword “full body portrait”. But more often than not it just doesn’t work. Below are a few examples with “full body portrait” keyword added. You can see the keywords are completely ignored.
A more consistent approach is to add keywords of what you want to see in the picture. For example
- long dress
Belows are generated with adding keywords standing and long dress.
Now it is showing more of the body but we can do better. The last trick I have is to change image size to a portrait size on top of using these keywords. Below are some examples. That’s all I know to tease full-body images out of the model!
Garbled faces and eyes
As detailed in a previous article, it is fairly common to see distorted faces. In fact it happens to other body parts, just that our brain is particularly sensitive to any asymmetry in faces.
To fix faces and eyes, turn on face restoration if your GUI supports it. See my previous post for detail. If you are using a free site with no built-in support for face restoration, you use the following sites for post-processing.
It’s well-known in stable diffusion community that the AI has problem getting fingers right. It is not uncommon to see distorted, too many or too few fingers.
The first fix is to include keywords that describe hands and fingers like “beautiful hands” and “detailed fingers”. That tends to prime the AI to include hands with good details.
The second fix is to use inpainting. Create a mask in the problematic area. Use inpainting to generate multiple images and choose the one you like. I found it sufficient to reuse the same prompt as the original image.
Thats all for today. Hope you find these techniques useful!
Follow me on Twitter to receive content like this.
Subscribe to my newsletter to receive weekly updates.
Support this free blog by using my products