Every photo cleanup request arrives in the same shape: a picture with too many people in it and a vague instruction to fix it. The person asking usually knows exactly who should go, but they describe it as a mood rather than a position, and the editor spends the next hour guessing. People Remover takes the opposite order: you write the sentence first, then the edit follows, because according to the product page the tool works from a plain description of who to remove instead of a hand drawn mask.

That ordering sounds trivial until you compare it with how masking actually goes. A brush selection is precise about pixels and silent about intent. If the brief says remove the person on the left and the photo has four people on the left, the mask cannot resolve the ambiguity, but a sentence can: the man in the red jacket on the left is a different instruction.

1. Write the sentence before you open the tool

The useful exercise is to describe the unwanted person the way a witness would. Position, clothing, action, and relative size cover most cases: the man on the far left, the woman in the red jacket, the child sitting on the steps, everyone in the background except the couple. According to the product page, that is also how the service expects to be addressed, with example instructions such as removing one person by position or clearing a crowd while keeping the main subject.

If the sentence has two possible readings, the edit will too. Rewrite before uploading rather than after the first result.

2. Where a description beats a selection

Brushing wins when the boundary is visible and geometric: a sign, a cable, a logo on a shirt. Language wins when the boundary is social rather than visual. Ask someone to mask the groom's ex partner's arm in a group photo and watch the conversation stall. Ask them to remove everyone except the bride and groom and the intent is unambiguous, which is why the description first workflow is easier to delegate to somebody who was not at the event.

3. The files people actually have

According to the product page, uploads accept JPG, PNG, WebP and iPhone HEIC or HEIF photos up to 10 MB, and HEIC files are converted to JPG in the browser so the format does not become the blocker. Downloads come back as JPG, PNG or WebP, with 2K and 4K output described as options on paid plans, and the page states there is no watermark on any plan. For a phone first workflow those details matter more than the model behind the edit: most people are working from a camera roll, not a layered file.

4. When a reshoot is still the right answer

Removal is a repair, not a redesign. If the person occupies a third of the frame, overlaps the subject's face, or carries a shadow that defines the composition, the honest option is another frame. The same applies when the removal changes what the photo claims happened. A crowd shot edited into an empty street is not a cleanup, it is a different statement, and that distinction belongs in a policy rather than in a retouching session.

Practical triage before uploading:

Small, partially occluded, background: a description and one pass is usually enough.

Mid frame, overlapping a prop: expect two passes and a check of the edges where the prop meets the body.

Foreground, touching the main subject: budget for manual finishing, or reshoot.

Anything that changes who was present: ask for written permission before editing, not after.

5. The check before you publish

Look at the result on a phone screen at arm's length, not at full resolution. Removed people leave traces that read as texture problems: a duplicated paving stone, a leg of a chair that no longer has an owner, a horizon that bends where a shoulder used to be. Then confirm the frame still supports the caption. If the text mentions a crowd and the image is now empty, the edit has created a factual problem that no amount of cleaning will fix.

The reason to start with wording is not that language is more advanced than pixels. It is that the sentence is the only part of the job a non editor can review. People Remover is a reasonable fit when that sentence is the input, the photo comes from a phone, and the person who wants the change is the person who will approve it. When any of those three is missing, slow down before uploading, because the tool will follow your description exactly, including the parts you did not mean.