How to Add a Person to a Photo: AI, Apps, and Realistic Results

AI generation, AI merge, and manual cutout: what changes

The technical approach determines how the composite is generated and how much control you keep over pixel integrity. Pure AI generation creates a new subject from a text prompt. AI merge tools, such as Magic Merge, use mask-guided inpainting to rebuild the target region while preserving surrounding context (OpenAI API Documentation, 2026). Manual cutout isolates pixels from a reference image with vector paths or selection brushes and generates no new visual data (W3C CSS Masking Module Level 1, 2021). Choose the method by priority: exact identity, or an integrated pose that fits the scene. Both goals rarely win at once.

If you want to compare the underlying feature sets of individual applications before committing to a method, review our reference material on AI photo editors and their core editing modules.

When adding a person from one photo works best

Adding a person from one photo to another works best when both sources share compatible capture conditions. High-quality composites rely on matching camera angles, uniform lighting direction, and comparable resolution between the reference portrait and the base scene (NIST SP 500-290e3, 2025). Research on affordance-aware insertion shows that near-frontal poses and uncompressed face detail let diffusion models preserve identity far more reliably during transfer (Kulal et al., "Putting People in Their Place," CVPR 2023).

Biometric capture standards give useful numeric anchors even for creative work. NIST best-practice portrait guidance places head width at roughly 50% of frame width, sets the eye line near the 55% vertical point, and limits roll, pitch, and yaw deviation to about ±5 degrees for high-quality face imagery (NIST mugshot best-practice guidance). A reference portrait that respects those ranges gives the model substantially more usable identity signal. Photos work best when nothing fights the model: no motion blur, no heavy filter, no aggressive JPEG artifacts.

Choose photos with matching light, angle, and image quality

Make sure the base photo and the person image share identical or at least complementary capture geometry. Mismatched light direction, hard contrast against soft diffusion, or different camera elevation angles reveal synthetic manipulation instantly (NARA Digitization Quality Management Guide, 2023). Standard image specifications recommend consistent aspect ratios and spatial resolution above minimum density thresholds, such as 20 pixels per foot for observation and 80 pixels per foot for recognition, which helps avoid focal softness around the added subject (DHS Digital Video Quality Handbook, 2024). NIST SP 500-290e3 defines aspect ratio as the width-to-height relationship of the captured image and cites 480x600 pixels at 1:1.25 as a workable biometric minimum.

Crop the person cleanly and preserve important details

Isolating a subject demands precise boundary selection, otherwise background spillover travels into the target image. Hard-edge cutouts work fine for solid clothing contours. Hair strands, sheer fabrics, and complex edges need refine-brush masking or soft-edge alpha matting (Cutout.Pro Technical Guide, 2026). An automated background remover with micro-feathering keeps subtle boundary pixels partially transparent, which prevents dark halos when the subject sits over a lighter background.

A practical edge taxonomy helps here. Treat solid clothing and footwear as hard edges. Treat hair, fur, and fluff as soft edges. Treat glass, lace, or sheer fabric as complex translucent edges. Each class needs its own refinement pass: binary masks for the first, feathered alpha for the second, and manual Keep/Remove brushwork at high zoom for the third. Most tools automatically detect the obvious silhouette; the last 5% of the boundary is still yours.

Upload the background photo and the person you want to insert

Open your chosen AI photo editor and upload the original photo that serves as the destination scene. Next, upload the person image that contains the subject you want to insert. Most platforms let you designate one image as the structural background and the second as the subject reference (EditThisPic Workflow Specifications, 2026). Simply upload the destination background first where the interface allows it. That single habit locks output resolution and aspect ratio before the subject layer gets scaled.

Technical tip for high-density group edits: with advanced AI merge tools, state the total target subject count in the settings or prompt text, for example "final output must contain exactly 5 people." Diffusion models use that numeric constraint to allocate attention-head maps more accurately. The result: fewer duplicated facial features, and far less risk of adjacent bodies merging into one malformed figure. Several dedicated group-merge services now expose this as a required input field before upload.

Write a prompt that defines placement and appearance

Write a structured prompt that directs subject position, pose, interaction, and context. Effective prompt engineering follows a strict order: scene context, subject identity, relative placement, then preservation constraints (OpenAI GPT Image Generation Prompting Guide, 2026). For edits, separate what must change from what must stay identical, namely identity, geometry, camera angle, layout, lighting, and surrounding objects. Repeat the preserve-list on every iteration. Drift is the default behaviour, not the exception.

• Example prompt: "Insert the young woman from the reference image standing on the right side of the group photo, wearing a navy blue blazer, looking toward the camera, smiling naturally, matching the soft evening outdoor lighting, keep all surrounding background elements unchanged."

When you run complex media transformations or integrate multi-stage asset pipelines, our AI Media Workflows directory documents the architectural patterns.

Ready-to-Use AI Prompt Templates for Specific Scenarios

Copy and adapt these prompt structures for your target scenario. Each one follows the scene, subject, placement, constraint order recommended by vendor prompting documentation.

• 1. Couple or friend portrait

  "Insert the person from the reference image standing naturally beside the main subject in the target photo. Match the background color temperature, skin-tone midtones, and soft directional light from the left. Retain original face proportions and expression. Keep the background, camera angle, and framing unchanged."

• 2. Adding a pet (dog or cat) to a family shot

  "Place the pet from the reference image on the floor next to the family members in the base photo. Match the carpet depth plane, cast a soft ambient ground shadow beneath the paws, align environmental lighting, and keep realistic body proportions and natural eye direction."

• 3. Inserting one person into a large group

  "Insert the reference person into the middle row of the group shot between [Subject A] and [Subject B]. Scale shoulders and head size to match the surrounding depth plane, align the eye line with adjacent faces, and apply matching camera depth-of-field blur. Final output must contain exactly [N] people."

• 4. Recreating memorial or legacy photos

  "Seamlessly integrate the individual from the reference photo into the family background. Harmonize film grain and age-related noise, match vintage contrast and exposure, and ensure exact horizon and eye-level alignment. Do not alter any existing faces."

• 5. Merging two separate group photos

  "Combine the people from the first group photo with those in the second photo. Keep the lighting, environment, and camera angle of the second photo unchanged. Arrange both groups harmoniously as if photographed together in a single frame."

• 6. Emotional memory portrait (reunion framing)

  "Blend the person from the first photo into the second photo beside the existing subject. Keep all lighting, pose, and atmosphere from the second photo unchanged. Ensure both faces and expressions look naturally aligned as if captured in the same moment."

Interactive prompt generator (widget specification)

To cut trial-and-error, the prompt builder embedded above the tool comparison table assembles a constraint-complete prompt from three dropdown selections.

The widget concatenates selections into the scene, subject, placement, preservation sequence, appends a numeric subject-count constraint, and exposes a single Copy Prompt action. The string pastes directly into Firefly, CapCut, Gemini, or an API call.

Generate, review, and download the edited photo

Start generation and inspect the candidates that come back. Advanced AI will automatically handle blending, but it also fails quietly, so check localized artifacts: distorted hands, unnatural shadows, edge blur, or shifted background elements (BBC Research Note on Synthetic Media Detection, 2025). Review order matters. First confirm prompt adherence. Then scan hands, limbs, faces, and any text for warping or merged geometry. Finally verify that nothing outside the masked region changed. Once the added person looks realistic and the lighting transitions stay smooth, download the final image as uncompressed PNG or high-quality WebP, since transparent output requires PNG or WebP in most APIs.

Troubleshooting: which artifact points to which mistake

Quality results usually come from one more iteration, not from a better tool. Use this table as a triage sheet before you re-generate.

Add someone with a cutout tool and manual positioning

Online editing tools support manual composites through plain drag-and-drop. Upload your reference image to an online background remover to generate a transparent PNG cutout (Canva AI Workflows, 2026). Import the clean cutout onto your target background photo, resize the figure with corner transformation handles, then adjust brightness and contrast to match the ambient room tone. Save as PNG when you need transparency preserved, or as high-quality JPG once the composite is flattened and final. Working with a photo online free of charge is realistic here, provided you check the export ceiling first.

Add a person on iPhone or Android with a photo editor app

Mobile apps on iOS and Android compress the whole workflow into a smartphone-friendly sequence. An app to add person to photo, such as YouCam Perfect or MyEdit, ships dedicated "Add Photo" and "Cutout" modules that isolate a portrait and place it over a destination picture (YouCam Perfect Mobile Guide, 2026). Two interaction models coexist on mobile. The conversational AI Agent mode takes a single sentence, for example "add this person to the group photo and make it look natural," and handles positioning and blending itself. The manual Photo Edit mode leaves the cutout and transform under your control. Pick the agent mode for speed, the manual mode when identity fidelity and exact placement matter.

If your goal is a polished portrait rather than a group composite, our guide to AI headshot generators covers background replacement presets built for professional imagery.

Step-by-step mobile workflow (iOS & Android)

A precise cutout placement on a smartphone needs no desktop software. Follow this sequence.

• Scale and rotation: pinch the corner handles until the subject's eye-line height matches adjacent subjects on the same depth plane.
• Target count hint: in an AI merge mode rather than manual layers, set the total expected subject count to [N] so the model rebuilds edges without duplicating faces.
• Orientation: mirror the cutout only when doing so does not reverse text, logos, or asymmetric facial features.

Quick reference, one line each. To combine two people in one photo for free, cut out each subject separately and merge them as two layers. To add a partner or a friend, use either an AI Replace brush with a descriptive prompt or the Cutout plus Add Photo path. To place a face onto another image, remove the background from the face crop first, then scale it to match the target head width before blending.

1. Launch the mobile editor and select the base scene.Open your app (YouCam Perfect, MyEdit, Canva Mobile, or Fotor Mobile) and tap Photo Edit. Load the destination background photo first so scene resolution and aspect ratio are locked.
2. Import the subject via the cutout module.Tap Add Photo and select the reference portrait. Trigger the automated AI Cutout / Background Remover immediately to isolate the subject boundary.
3. Adjust the control parameters.
4. Refine edges and integrate shadows.Tap Refine Edge to micro-feather hair boundaries at high zoom, then paint a roughly 15% opacity dark radial shadow beneath the shoes to anchor the subject to the ground plane.
5. Export.Save to the camera roll in PNG or uncompressed JPEG. Verify the export resolution, because several mobile apps silently downscale free-tier saves.

Match lighting, shadows, color, and skin tone

Mismatched lighting is the number one tell of a synthetic edit. Inspect the base scene to identify primary light sources, watching highlight placement and shadow cast directions on existing subjects. Updated note: the widely cited environment-harmony metrics from SwapAnyone (2025) were developed and validated on video sequences, so treat them as directional guidance for stills rather than a validated single-frame standard. For still compositing, the classical checks stay more defensible: matched highlight, midtone, and shadow values, plus consistent gamma. Adjust color balance with a vectorscope or RGB channel ratios so skin tone midtones agree with ambient environmental reflections. Then cast a drop shadow beneath the feet or the seating surface with a low-opacity dark brush, which anchors the subject to the ground plane.

Practical lighting checklist, in this order:

1. Identify the key light vector.Trace an existing subject's shadow back to its source and note elevation and azimuth.
2. Rotate or re-light the cutoutso its highlight side faces the same azimuth. If the reference portrait is lit from the opposite side, replace the reference. Repainting directional light convincingly is harder than finding a better source photo.
3. Paint the contact shadow.Use a small, hard-edged dark brush at 25% to 40% opacity directly under the footwear or seat contact points. This is the shadow that sells ground contact.
4. Paint the cast shadow.Add a longer, softer, lower-opacity shadow extending away from the light source at the same angle as existing shadows, then blur it progressively along its length to imitate penumbra falloff.
5. Unify color.Sample a neutral background midtone and apply a low-opacity grade layer clipped to the subject, so ambient bounce light matches.
6. Match noise and grain.Add grain to the inserted subject until its texture matches the base image at 100% zoom.
7. Match edge softness.Apply a one-to-two-pixel blur to the subject boundary when the base image came from a softer lens or a higher ISO.

A decent photo enhancer pass at the end can unify sharpening across the frame. Run it on the merged file, never on the cutout alone.

Correct perspective, scale, and placement in a group shot

For natural perspective in a group shot, align the added subject's eye line and head scale with people on the same depth plane (Landscape Institute Photomontage Guidance, 2025). A figure placed off the horizon line, or scaled larger than nearby participants, breaks optical consistency at a glance. Keep faces at roughly equal scale across rows. Never let the new subject obscure an existing face. Make sure distance blur and camera depth-of-field affect the added person exactly as they affect surrounding scene elements. Automatic perspective correction presets help, though they do not know which row you intended.

If the composite needs extra canvas around the group after repositioning, for banner crops or print bleed, our overview of AI outpainting tools for expanding images explains how to extend backgrounds without distorting the newly inserted subject.

What "free" includes: credits, exports, and watermarks

Free AI photo editing tools usually structure access around recurring credit allocations, capped output resolutions, or watermarked exports (Cutout.Pro Tier Analysis, 2026). A service may be free to use and still charge credits for a full-resolution, uncompressed PNG. Some vendors bundle all three constraints at once: a starter plan with a few hundred credits, a 576-pixel output ceiling, a visible watermark, and an explicit non-commercial clause. Others delete uploaded originals after processing yet keep a perpetual licence over generated outputs. Read credit renewal intervals, export caps, and licence scope before large projects, and cross-check our roundup of free AI art generators for current limits.

Features to compare before choosing an AI photo editor

When comparing editors, prioritize features that give explicit control over placement and context matching. The essentials: automatic background removal, mask-based inpainting, perspective correction presets, refine-edge brushes, generative expand, generative upscale, and multiple people support (Adobe Firefly Feature Matrix, 2026). Flexible customization beats a single clever button. For structured performance evaluations across media creation tools, consult our AI Media Comparison Matrices.

If you also need to check whether an incoming image has already been synthetically altered, our material on AI reverse-image-search and detection tools covers provenance checks and detector accuracy.

Complete a family photo or group photo without a retake

Inserting a missing family member removes the cost of a reshoot. Mask the open space in the base group shot, supply a clear reference portrait, and AI diffusion models can blend the individual into the existing lineup (GroupDiff ECCV Study, 2024). Render lighting and shadow contact points beneath the footwear carefully, otherwise the added person floats. The same workflow serves corporate and event contexts: updating a team photo after a new hire, completing a graduation series, or reuniting relatives separated by travel restrictions and scheduling conflicts. A family gathering where one person was stuck at work is the most common request we see, followed by wedding photo edits.

For teams that publish these composites as video recaps or social reels, our guide to YouTube video editors and publishing workflows covers export settings that preserve composite detail through platform re-encoding.

Creative, lifestyle, and social media applications

• Self-cloning and time-lapse effects. Duplicate your portrait across several positions in one room or landscape to tell a visual story, show one subject in multiple actions, or build a surreal solo party group shot. Repeat the cutout-and-place cycle, varying pose and scale per instance, so each clone sits on a plausible depth plane.
• Virtual travel and backdrop swaps. Isolate your portrait and place yourself into landmark environments: museum interiors, coastlines, city squares. Auto-match ambient bounce light and horizon height. This is also the fastest way to preview whether an outfit reads well against a specific wedding or photoshoot location.
• Adding pets and objects to a scene. Insert cats, dogs, bags, hats, or accessories using the same perspective-vector and contact-shadow rules that govern human portraits. Pets need extra attention on ground shadow and eye direction, since floating paws are the most obvious tell.
• Outfit previews and social stickers. Cut out individual fashion items or full-body poses to build custom sticker layers for style moodboards, outfit-of-the-week grids, and product-styling previews.
• Memes, pranks, and entertainment edits. Place a friend's cutout into an absurd setting or merge multiple photos into a single gag composite. Keep these clearly playful and confined to private chats unless you hold explicit consent to publish.
• Marketing, product, and team visuals. Designers and marketers insert staff, models, or clients into brand visuals without scheduling a shoot, matching studio lighting and perspective for campaign banners, landing pages, and slide decks.
• Channel art and creator assets. Composites travel well into channel branding: a cutout portrait becomes a youtube pfp or feeds a youtube pfp maker template, the wide version feeds a youtube banner creator layout, and the same merged frame can open a youtube intro built in a youtube intro maker or close a video through a youtube outro template. Export at 2x the target canvas so re-crops do not soften the inserted subject.

Create memorial, social media, and professional images responsibly

Creating memorial pictures of a deceased loved one, or editing public social media shots, demands genuine care. Composites of this kind, a grandparent placed beside a grandchild they never met, a relative added to a wedding they could not attend, carry real emotional weight. That is exactly why consent and labelling requirements matter. For marketing or commercial use, confirm that every depicted living individual has granted explicit consent (EU AI Act Transparency Guidelines, 2026). Consent is purpose-specific: permission for a family album does not transfer to an advertising campaign. Creators must also evaluate copyright terms governing commercial rights. For detailed licensing analysis, consult our guide on commercial use rights across AI-generated media and our breakdown of commercial-use rights for AI-generated images.

This information is general in nature and does not replace legal advice on copyright, data protection, or personality and image rights. Requirements vary by jurisdiction.

Internal workflows and hub directory

To explore our full repository of automated asset pipelines, production templates, and media processing documentation, visit our primary AI Media Workflows hub.