Inpainting: what it is and how it edits only part of an image
Quick facts
- Definition
- A technique in which AI fills in or redraws only a chosen area of the image, keeping the rest unchanged.
- Contrast
- Instead of generating a whole new image, it changes one section.
- Use in HOF
- Simulating change in areas of the face while keeping eyes, hair and background.
- Limit
- The edited area may end up with a texture or contour different from the real one.
Inpainting is a generative AI technique in which only part of the image is redrawn. The area is indicated by a mask, and the model generates content that matches what surrounds it. Everything else in the image is preserved.
Why does it matter for face projections?
A good projection should change only what was planned. If the model regenerated the whole face, it could change the gaze, expression, skin or hair, and the comparison would lose its meaning. By restricting the edit to the chosen area, the "after" image stays closer to the "before", and the patient clearly sees only the planned difference.
| Approach | What happens |
|---|---|
| Generating the whole image | The model may change traits nobody asked for |
| Inpainting | Only the marked area is redrawn |
| Text instruction | The model follows the instruction but may overreach |
What are the limits?
The redrawn area may have a slightly different texture or contour. The model can also "invent" details, a subject covered in hallucination in image AI. That is why any output is an illustrative simulation and must be presented as such.
Frequently asked questions
What is inpainting in an image?
It is the editing of a specific area of an image by AI, which fills the area with content consistent with its surroundings and keeps the rest of the image as it was.