Verified feature
Portrait backgrounds
Detect a portrait subject on the device and replace the surrounding background.
The problem it solves: A portrait may need a cleaner or differently colored background while keeping the person visible.
What you can do
- Run on-device portrait subject detection.
- Refine the detected portrait mask.
- Replace the surrounding area with a selected background color.
- Preview the composition before saving.
Available options
- Background color
- Portrait composition and preview
Supported inputs
- A portrait image with a detectable person
Outputs
- Portrait image with a replaced background
Workflow
- Choose a portrait with a clearly visible subject.
- Let the on-device model detect the person.
- Inspect the subject edge and composition.
- Choose the replacement background.
- Generate and review the output.
Limitations
- Edge quality depends on the source image, lighting, pose, and model result.
- The first use can require downloading the portrait model.
Common questions
Is portrait detection performed on the device?
The verified workflow uses an on-device portrait model after its required model asset is available.
Will every edge be detected perfectly?
No. Hair, low contrast, complex backgrounds, and image quality can affect the generated mask, so preview review is required.
Project evidence
feature/feature_app/src/main/java/com/wkq/photo/feature/app/aibackground/AiBackgroundFragment.ktfeature/feature_app/src/main/res/values/home_strings.xml