Freepik AI Inpaint: Fix Faces in Group Photos Without Photoshop
Learn how to remove photobombers, adjust expressions and balance lighting in group photos with Freepik AI Inpaint.

👁 114 reads
Group photos are a photographer’s nightmare. Thirty people lined up, and someone is still squinting, someone else is half out of frame, and one person has sneezed at exactly the wrong moment. Until recently, the only option was Photoshop, hours at a monitor and a hope that your layer mask turned out right. Freepik AI Inpaint takes a different approach: select a face, describe what you want to change and wait a few seconds.
This tutorial walks you through three practical photography scenarios: removing an unwanted person from a group shot, changing a facial expression and evening out the lighting across faces. Each scenario includes a prompt you can copy and an explanation of why that wording works. No waffle or theory - just practical steps.
What you need before you start
You need a Freepik account with an active premium plan - AI Inpaint isn’t available on the free plan. Plans start at several dozen Polish zlotys per month, making it a relatively small expense for a photographer doing a dozen or more shoots a month. Prepare your photo first: use an image with a resolution of at least 1,200 pixels on its shorter side, in JPG or PNG format. RAW files won’t work, so convert them in Lightroom or your camera’s software before you start.
After signing in, find AI Inpaint under AI Tools, then Image Editor. It runs in your browser, so there’s nothing to install. Upload your photo, select the brush tool, paint a mask over the face or person you want to change, enter a prompt and click Generate. That’s the theory. Now for the practical examples.
Tip
Before uploading a photo to Freepik AI Inpaint, reduce its file size to no more than 4-5 MB. Larger files slow down generation without improving the result - the model processes the image at its own internal resolution anyway. Use TinyPNG or export from Lightroom at 85% quality.
Scenario 1: Remove a photobomber from a team photo
It’s a familiar situation: the marketing team photo looks perfect, but a pizza delivery person has wandered into the frame at the left edge. Masking them out in Photoshop could take half an hour with the Clone Stamp and Content-Aware Fill. With Freepik AI Inpaint, it takes about three minutes.
Use the brush to select the entire unwanted person - not just their face, but their whole body, with a small margin of about 20-30 pixels around it. A mask that’s too tight can leave artifacts along the edges. Then enter a prompt describing the background that should appear where the person was:
empty office interior background, matching the surrounding area, natural lighting, no people, seamless continuation of the background, professional corporate environment
This prompt works because it describes what should appear after the edit, rather than what should be removed. The model gets a clear instruction: an office background, natural lighting and no people. “Seamless” is key because it directs the model’s attention to the mask edges.
empty outdoor park background, matching ambient lighting, blurred green trees and grass in background, bokeh effect, no people visible, continuation of existing background
This version is for outdoor photos - for example, if you want to remove someone from a group photo in a park. Replace “park” with a suitable setting, such as “beach,” “conference room” or “garden.”
Tip
If the person you’re removing is standing next to someone and their bodies touch, split the edit into two masks. First, select and remove most of the person, leaving the area where the two figures meet. Then use a second pass to refine the edge next to the other person. In this situation, one large mask will almost always distort the neighbouring person’s face or arm.
Scenario 2: Change a facial expression
A common problem in wedding and corporate photos: everyone is smiling, but one person looks as if they’ve just heard about an audit. Or the reverse - the photo is serious and formal, but someone is sticking their tongue out at the camera. Freepik AI Inpaint can change a facial expression without changing the rest of the head.
Mask precision determines the result. Select only the face - from the hairline to the chin and from ear to ear. Don’t include the neck or hair. The more precise the mask, the more natural the transition. Then try this prompt:
natural warm smile, teeth slightly visible, relaxed facial expression, matching the original skin tone and lighting, photorealistic, same person, preserve hair and background
Notice “same person” and “preserve hair and background.” These phrases anchor the edit, telling the model not to invent a new face from scratch but to modify the existing one. Without them, you might get perfect teeth and a smile that belong to someone else.
serious and composed expression, mouth closed, professional look, neutral facial expression, photorealistic, matching original skin tone, same facial features, preserve surrounding elements
Use this version for formal team photos and institutional portraits. “Composed” works better than “serious” on its own: it helps the model understand that the face should look calm, not sad.
eyes wide open, alert and attentive expression, natural look, same eye color, photorealistic, matching original lighting conditions, preserve surrounding facial features
This prompt can rescue a photo where someone blinked at the wrong moment. It works best when closed eyes are the only problem and the rest of the face is sharp and well lit.
Note
When changing a facial expression, Freepik AI Inpaint sometimes “corrects” the person’s age too, particularly for older faces with distinctive wrinkles. If the result looks overly smooth, add “preserve natural skin texture, realistic skin details, maintain age-appropriate appearance” to your prompt. This won’t eliminate the problem completely, but it can reduce it considerably.
Scenario 3: Balance lighting across faces
Picture a group photo in a banquet room: half the people are standing under a chandelier with overexposed faces, while the other half are in shadow. Lightroom can adjust the overall exposure, but it won’t fix the lighting on each face individually. That’s where Inpaint comes in.
Select the face with the lighting problem - an overexposed face where details are disappearing, for example. Then use this prompt:
face with balanced natural lighting, soft even illumination, preserve facial features and expression, reduce harsh highlights, match ambient warm indoor lighting of the scene, photorealistic
face with corrected exposure, fill light added to shadowed areas, soft natural indoor lighting, preserve skin tone and facial expression, photorealistic, no harsh shadows under eyes or chin
This second version is for underexposed faces - a common problem in photos taken from the side of a room when the main light source is on the other side. “Fill light” is a photography term that the model understands and applies naturally.
Tip
When evening out the lighting across several faces, work from the worst case to the best. Start with the face that has the biggest exposure difference, generate and save the result, then upload the edited photo and work on the next face. Trying to fix every face at once with one mask can end in chaos: the model gets conflicting signals from areas with different lighting.
How to write prompts that keep the model on track
After dozens of tests with Freepik AI Inpaint, a few useful lessons emerge. First, describe the end result, not the action. The model isn’t a filmmaker - don’t tell it to “fix” something; tell it what should be visible. “Natural smile” works better than “make him smile.”
Second, always add technical quality anchors. These four phrases can improve prompts for photos of real people:
photorealistic, preserve surrounding elements, matching original lighting, same facial features
You can add them to the end of almost any face-editing prompt to help the result fit in with the rest of the photo. They’re not magic, but they give the model a clear signal: don’t invent; make an edit.
Third, use negative descriptions carefully. Freepik AI Inpaint has a separate “negative prompt” field. Try entering:
blurry, distorted, unrealistic, plastic skin, overly smooth, painting, illustration, artifacts, different person, changed background
A negative prompt lists things you want the model to avoid. “Different person” and “changed background” are particularly useful when editing faces. Without them, the model may replace the head or generate a completely new background behind the person.
Tip
Freepik AI Inpaint usually generates 2-4 variations at a time. Don’t pick the first one that looks acceptable - compare them all for mask-edge quality and how well the skin matches the rest of the photo. The best option isn’t necessarily the one with the most attractive face; it’s the one whose edges blend in best.
Common mistakes and how to avoid them
Using a mask that’s too small is the main mistake with this technique. If you select only the eyes to fix a blink, the model will generate new eyes that look as if they came from a different photo. Always leave a margin: the mask should be a little larger than the area you want to change.
Another common mistake is trying to fix a photo with too little resolution. If a face in a group photo is only 80 by 60 pixels, Freepik AI Inpaint can’t conjure details that aren’t there. A useful minimum face size is about 200 by 150 pixels. Below that, results may look blurry or strange. If you’re working with a phone photo where the face is tiny, first upscale the whole image with another tool, such as Topaz Gigapixel or Freepik’s upscaling feature, then edit it.
A third mistake is expecting too much when faces are blurred. AI Inpaint isn’t a reconstruction tool: if a face is unclear because the person was moving, you can replace it, but you can’t restore it. The result will be a new face, not the same face sharpened. For motion blur, use Topaz DeNoise or Photo AI.
Google · Your sourcesQuantslant higher in your Google - one clickAdd to preferred sources →Common mistake
Don’t upload photos with watermarks or company logos in the frame to Freepik AI Inpaint, especially if the mask touches the logo. The model often “fixes” logos by generating random text or distorted letters. Remove watermarks first, then edit the faces.
Should you give up Photoshop for good?
The short answer is no. The longer answer is: it depends on what you do.
Freepik AI Inpaint works well for quickly fixing one or two faces when the photo is destined for social media, a corporate presentation or a client’s website. It can cut the editing time from 30-45 minutes in Photoshop to 3-5 minutes in a browser. That’s a significant difference when you have dozens of photos from a shoot.
Photoshop has a clear advantage for large-format print, projects that need pixel-by-pixel control and any work where a client is paying for precise retouching. AI Inpaint has its good moments and its serious missteps. For now, a person with Photoshop can spot the latter better than any model.
The best middle ground is to use both: let Freepik AI Inpaint handle the first round of edits on 80% of the photos from a shoot, then use Photoshop on the 20% that need surgical precision. This division of work can cut post-production time in half without sacrificing the quality photographers care about.
More on this from QUANTSLANT
news·ChatGPT Ads Target Emails, Phone Numbers and PromptsA 404 Media test found that ChatGPT advertisers can upload contact lists and use conversation topics to guide ad targeting.

