# Remove objects from a photo with AI: people, wires, text and clutter — tested on one shot

> AI can remove an object from a photo with a text command — “remove the tourists on the left” — no brush, no masks. We took one shot with six unwanted objects and removed them in Nano Banana Pro one by one and all at once — first try every time, with the face and light untouched. Then we compared Nano Banana Pro, Nano Banana 2.1 and GPT Image 2.5 on the same command.

Published 2026-10-14 · Updated 2026-10-14
URL: https://ai2face.com/blog/udalit-obekt-s-foto-nejroset

AI can remove an object from a photo with a single sentence: you write “remove the bin and the scooter on the right”, and the model finds them, erases them and redraws what should be behind them. Not long ago you had to brush over the object and clone the background by hand. We tested how well this works on one travel shot where everything gets in the way at once. All images below were made in the AI2Face generator; the person in the photo was generated for this test.

## The source: what we'll remove

![A man in a light blue linen shirt on a seaside embankment at sunset; two tourists with a selfie stick on the left, wires above, a bin and a scooter on the right, graffiti and a date stamp at the bottom](https://ai2face.com/_astro/before.BgsRvN2f_nVtqx.jpg "Source: six unwanted objects in one shot. The person was generated for this test")

The shot has everything you usually want to remove from a vacation photo:

| Object | Why it's hard |
|---|---|
| Two passers-by with a selfie stick | behind them are a wall, houses and cobblestones in perspective |
| Wires above the houses | thin lines against the sky and facades |
| Graffiti on the stone | the stone texture has to be rebuilt |
| Date stamp in the corner | sits on top of the cobblestones |
| Bin with a shadow | behind it are a railing and the sea; the shadow falls on the cobblestones |
| Scooter at the edge | cut off by the frame, with a wall and railing behind it |

## How to remove an object from a photo with AI: the prompt formula

Photo editing models understand plain language. A good removal command has three parts:

1. **What to remove and where:** “two tourists on the left, one with a selfie stick”.
2. **What should fill the space:** “continue the wall, houses and cobblestones”.
3. **What not to touch:** “keep the man, his face, pose, shirt and sunglasses, the light and colors exactly as they are”.

The third part matters more than it seems: without it, the model may “helpfully” adjust the face, light or colors too.

## 1. Remove people from a photo

```
Remove the two tourists walking on the left side behind the man (including the
selfie stick) and fill the area naturally with the continuation of the stone
wall, houses and cobblestones. Keep everything else exactly the same: the man,
his face, pose, shirt and sunglasses, the light, colors, wires, trash bin,
scooter, graffiti and the date stamp.
```

![Before and after: left, two tourists with a selfie stick behind the man; right, empty cobblestones and a wall in their place](https://ai2face.com/_astro/people.r815zzTt_cHr22.jpg "Left: source; right: no passers-by — the wall and cobblestones continue in perspective. Nano Banana Pro")

This is the most common request — removing random people from a photo at a landmark. The AI rebuilt the wall and the shadows on the cobblestones so that no seam shows, even zoomed in. The other unwanted objects stayed in place, as requested.

## 2. Remove wires

```
Remove all black overhead electric wires crossing the sky and the houses, and the
cable running down the wall on the right. Restore clean sky and facades in their
place. Keep everything else exactly the same.
```

![Before and after: left, black wires sagging over the houses; right, a clean sky and facades, the cable on the right wall gone](https://ai2face.com/_astro/wires.D94XY4iQ_Z102bHi.jpg "Wires and the wall cable removed, no traces on the sky or facades. Nano Banana Pro")

Wires are a classic retouching task: thin lines cross both the sky and walls of different colors. The model removed them along with the brackets and the box on the wall — and also the drainpipe, which we didn't ask for. If you need the pipe, add “don't touch the drainpipe”.

## 3. Remove text from a photo

```
Remove the red graffiti scribble from the stone block at the bottom left and the
orange date stamp in the bottom right corner. Restore the natural stone texture
and cobblestones in their place. Keep everything else exactly the same.
```

![Before and after: left, red graffiti on the stone and a date stamp in the corner; right, clean stone and cobblestones](https://ai2face.com/_astro/text.B76HA98I_ZhcpsI.jpg "The graffiti and the date stamp are gone, the stone texture is restored. Nano Banana Pro")

The same approach removes writing on walls, signs and date stamps from old cameras. We don't recommend removing other people's watermarks from protected images: that infringes the author's rights. If you want to replace text rather than remove it, see the text replacement test in our [AI photo editor](/blog/izmenit-foto-nejroset) guide.

## 4. Remove an object together with its shadow

```
Remove the grey trash bin together with its shadow on the cobblestones, and the
white scooter at the right edge. Restore the railing, the sea view, the wall and
the cobblestones behind them naturally, with consistent golden hour light. Keep
everything else exactly the same.
```

![Before and after: left, a bin with a long shadow and a scooter at the edge; right, a railing, the sea and empty cobblestones in their place](https://ai2face.com/_astro/bin.DTNv5R15_Z12gLPy.jpg "The bin is gone along with its shadow, and the railing and sea are rebuilt behind it. Nano Banana Pro")

A shadow is the telltale sign of a bad removal: the object is gone but its shadow is still there. Write “together with its shadow” explicitly. Note the railing: the model continued its even rhythm where the bin used to cover it.

## 5. Remove everything unwanted at once

```
Clean up this travel photo: remove the two tourists on the left (with the selfie
stick), all black overhead electric wires in the sky, the grey trash bin and its
shadow, the red graffiti on the stone wall, the white scooter at the right edge
and the orange date stamp in the bottom right corner. Fill every area naturally
with matching sky, houses, wall, sea, railing and cobblestones. Keep the man
exactly the same: face, sunglasses, shirt, pose. Keep the golden hour light,
colors and composition unchanged.
```

![Before and after: left, the shot with tourists, wires, a bin, graffiti, a scooter and a date; right, only the man on a clean seaside embankment](https://ai2face.com/_astro/all.DjpLrohX_Zk4sJe.jpg "Six objects with one command. The cable on the right wall stayed — this prompt didn't mention it")

The model removed six objects in one edit. The only thing left is the cable on the right wall: in this prompt we asked to remove wires “in the sky”, and the model took it literally. The lesson is simple: if an object matters, name it separately.

## Which AI is best at removing objects: three models compared

We gave the “remove everything” command to three models: [Nano Banana Pro](/models/nano-banana-pro), [Nano Banana 2.1](/models/nano-banana-2-1) and [GPT Image 2.5](/models/gpt-image).

![Three results of one command: left Nano Banana Pro, center Nano Banana 2.1, right GPT Image 2.5 with a brighter sky and a pronounced sun](https://ai2face.com/_astro/models-compare.BREOWDXj_Z1Uasy4.jpg "Left: Nano Banana Pro, center: Nano Banana 2.1, right: GPT Image 2.5")

| | Nano Banana Pro | Nano Banana 2.1 | [GPT Image 2.5](/models/gpt-image-2-5) |
|---|---|---|---|
| All 6 objects removed | yes | yes | yes |
| Person unchanged | yes | yes | yes |
| Light and sky as in the source | yes | yes | no — the sky got brighter and a sun appeared |
| Seams and artifacts | none visible | none visible | none visible |
| Best for | edits where precision matters | quick batch edits | when it's fine to “improve” the whole shot |

All three models handled the removal. The difference is discipline: Nano Banana Pro and Nano Banana 2.1 changed only what we asked for, while GPT Image 2.5 also redrew the sky and the sun. For a clean removal without side effects, use Nano Banana Pro or Nano Banana 2.1.

## Limitations: where AI can get it wrong

- **Literal reading.** “Wires in the sky” isn't the same as “a cable on the wall”. List objects specifically.
- **An object covering the subject.** If a passer-by hides part of the face or body, the model has to invent what it never saw — the result may differ from reality.
- **Many similar objects.** “Remove the third car from the left” works worse than “remove the red car”. Describe the color and position.
- **Repeated edits.** Each new edit is a new generation. If you need to remove five objects, do it in one command rather than five in a row: fewer chances to accumulate distortions.

If the photo needs a quality boost rather than a cleanup — removing noise and blur — see our guide to [AI photo enhancement](/blog/uluchshit-kachestvo-foto-nejroset). For old scratched photos, see [AI photo restoration](/blog/restavraciya-foto-nejroset).

## Where to remove objects from a photo with AI

Nano Banana Pro, Nano Banana 2.1 and GPT Image 2.5 are all in one AI2Face generator: upload the photo and try the command in several models. The cost of each edit is shown before you start, and credits from packs don't expire. A clean product photo on a plain background for a marketplace works the same way — see our guide to [AI product cards](/blog/kartochki-tovarov-nejroset).

## Bottom line

AI removes objects from photos without brushes or masks, in one sentence: in our test, Nano Banana Pro removed passers-by, wires, graffiti, a bin with its shadow, a scooter and a date stamp — one at a time and all at once, on the first try every time, without changing the face or the light. For precise edits, use Nano Banana Pro or Nano Banana 2.1; GPT Image 2.5 when you don't mind the model “improving” the whole shot. Name each object separately, ask to remove it “together with its shadow” and always add “keep everything else as is”.

## FAQ

### How do I remove an object from a photo with AI?

Upload the photo to the generator, pick an editing model such as Nano Banana Pro and write what to remove and what should fill the space: “remove the two tourists on the left, continue the wall and cobblestones in their place, keep everything else as is”. You get a clean shot in 20–40 seconds.

### Can AI remove a person from a photo?

Yes. Describe where the person stands and what they're wearing, and the model removes them and rebuilds the background. In our test, the AI removed two passers-by with a selfie stick on the first try without touching the main subject.

### Do I need to brush over the object?

No. Text-based editing models understand “the bin on the right”, “the wires in the sky”, “the writing on the wall”. A brush only helps when there are many similar objects and you need to remove one specific one.

### Does AI remove the object's shadow too?

Yes, if you ask: “remove the bin together with its shadow on the cobblestones”. Without that phrase, the shadow can stay behind on the empty spot.

### Can I remove objects from a photo for free?

Yes, a few times. AI2Face gives you starter credits when you sign up, and one edit is inexpensive — the cost is shown before you generate.

## Sources

- https://blog.google/innovation-and-ai/products/nano-banana-pro/
- https://ai.google.dev/gemini-api/docs/image-generation
- https://developers.openai.com/api/docs/guides/image-generation