Ideal inputs
Upload one image and describe the target region precisely, including its position, approximate size, and the visual structure that should remain around it.
Focused AI image workflow
Inpaint
Replace a described region of an image by asking the surrounding light, texture, and perspective to continue through the edit.
How to use it
1
Upload the source
Choose a clear image with the area to change visible and enough surrounding context for the model to follow.
2
Describe the replacement
Name what should appear in the region and specify nearby texture, lighting, perspective, and edges to preserve.
3
Inspect the boundary
Check the blend, shadows, repeated patterns, object edges, and any content just outside the described area.
Useful for
- Explore a replacement detail in a product image
- Test a set-extension or prop direction
- Repair a small visual area in a concept image
Expected result
GPT Image 2 creates a localized edit concept based on the described region and its surrounding image context.
Limitations to review
This page does not expose a dedicated mask or brush. Text-only region descriptions can be imprecise, especially on busy scenes or fine boundaries.
Frequently asked questions
Is there a manual mask tool?
No. Describe the region with position and nearby visual landmarks in the prompt; review the result because the boundary is inferred.
Can it preserve everything outside the region?
The workflow asks for that, but image edits can affect nearby pixels. Compare the surrounding edges and subject details with the source.