AI keeps replacing the person you want to remove? Try this workflow

First decide whether the problem is targeting or replacement

A new person appearing in the old person’s place is a different failure from removing the wrong person. In the first case, the edit has changed the figure but retained a person-shaped subject. In the second, the request or selected region may have targeted someone else. Diagnose the visible result before changing every part of the prompt.

You cannot establish the model’s internal reason by looking at one output. An incomplete target, ambiguous wording or a difficult overlap may contribute, but none is a universal explanation. What you can control is the original photograph, the scope of the request and a description of the intended empty space.

Start again from the source if the result has introduced a new face or damaged the foreground subject. For the complete process, see removing people from photos.

Describe the person and the space they should leave

Use two identifiers that actually appear in the picture: clothing and location, for example. “The passerby in a red coat at the far left” is less ambiguous than “the person.” Then describe the replacement surface. An uninterrupted stone wall gives the editor a more concrete task than a vague request to improve the photo.

Do not ask to replace, transform or restyle the unwanted person. Those words describe a different goal. If an area-selection control is available, include the whole visible figure rather than only the face or torso. Controls vary by version; use the prompt to state the desired result even when you also select an area.

Check the pieces that belong to the unwanted person

A person can occupy more of a scene than the body outline suggests. A bag, a shadow, a reflection in a window or a partly visible foot can remain after an otherwise good removal. Include those details when they should disappear, but distinguish them from nearby belongings that belong to your subject.

Follow the figure from head to feet in the original. If the person continues beyond the edge of a crop, consider using the uncropped original when you have it. If their arm crosses your subject’s shoulder, recognize that the missing shoulder detail will have to be inferred. A crop or another frame may preserve the memory better.

• Look for reflections in windows, mirrors, water and polished surfaces.
• Identify the direction and extent of the person’s shadow.
• Check bags, straps and loose clothing around the figure.
• Keep unrelated foreground objects out of the removal request.

Use a short retry plan in AIPGEN

Choose the original in AIPGEN and work on one unwanted figure first. The public listings support removing people; they do not promise that every crowded scene can be reconstructed accurately. Keep the first attempt simple enough that you can tell whether it succeeded.

1. Identify the unwanted person by position and clothing.
2. State the empty surface or background that should replace them.
3. Name the main subject and important details to retain.
4. Generate and compare the whole photograph with the original.
5. If another person appears, return to the original and simplify the request.
6. Save a separate copy only after checking the foreground subject and repaired area.

Handle a crowd differently from one photobomber

A crowd behind a portrait is not just many copies of a simple removal. Each figure may hide architecture, overlap another person or cast a different shadow. Decide whether the purpose is to remove one distraction or to create an imagined empty location.

For one conspicuous photobomber, remove that figure and leave the rest of the scene alone. For a heavily obscured background, a new background may be more honest as a creative interpretation. Use the background-change guide if that is the result you actually want.

If a person covers your subject’s face or hand, the generated replacement cannot recover a verified record of the hidden detail. Prefer another photograph when accuracy matters. Explain substantial creative changes when you share an image that could otherwise imply a different event.

Review the result before trying again

Check the empty space, then check what should have stayed unchanged. Follow straight lines through the repaired region: paving joints, railings and wall edges should connect. Inspect faces, clothing and the outline of the main subject separately.

If the person is gone but the area looks smeared, the problem has changed. Use the blurry-patch repair workflow rather than repeating person-removal instructions. For other failures, the AI editing troubleshooting guide helps separate targeting, structure and appearance errors.

Stop when the photograph works at its intended viewing size. A small crop is sometimes better than more generations. AIPGEN’s complimentary edit is limited; review the current paid offer before making repeated attempts.