What is Multiple Face Swap?

Multiple Face Swap is an online group-photo tool that detects and numbers up to five faces. You can replace one person, map different source faces to selected people, or use one source face across the group while leaving unselected faces unchanged.

How does the face detection work?

The same provider used for generation scans the group photo and returns a cropped preview for each destination index. The page keeps those stable face numbers in a short-lived server session so you can map replacements without guessing positions.

Can I replace just one person's face?

Yes. Select one detected face number, upload one replacement portrait, and leave every other face unchecked. Only the selected target is included in the replacement map.

What if I want to change everyone's face?

Choose the same-face mode, upload one source portrait, and select every detected person you want to replace. The server expands that source into the selected target mappings.

How do I help the tool detect faces accurately?

Use a clear group photo with no more than five visible faces. Avoid tiny, heavily blurred, cropped, or obstructed faces, and use replacement portraits with one clearly visible face and a reasonably similar angle.

What does mapping mean in Multiple Face Swap?

Mapping defines which uploaded source portrait belongs to which detected target number. Review every source-to-target pair before generation so the intended person receives the intended replacement.

Is my group photo private while using the tool?

Detection previews use short-lived signed links instead of public object URLs. Inputs and results remain private to the processing workflow, subject to the site's storage and retention controls.

How do I use Multiple Face Swap for a group photo?

Upload a JPG, PNG, or WebP group photo and confirm consent. Let the tool detect and number the faces, choose a mapping mode, add the required replacement portraits, review the relationships, then generate and download the result.

Do I need credits to use Multiple Face Swap?

A signed-in multiple face swap uses 2 credits. Credits are refunded once after a clear provider failure. A first eligible guest may use the site's shared one-time AI face swap trial.

What should I do if the face swap does not look right?

Check that every selected number is mapped to the intended source, then try clearer replacement portraits with similar angles and lighting. A target face that is blurred, hidden, very small, or strongly angled may need a different group photo.

Detect up to 5 faces 2 credits Server-verified map

Multiple Face Swap Online

Detect every face in one group photo, see its number, and control the relationship between each person and each replacement image.

Multiple Face Swap workspace

Detect, number, map, and replace up to 5 faces · 2 credits

Unmapped faces stay unchanged

Upload a group photoDrag & drop or choose a clear JPG, PNG, or WebP with no more than five visible faces.I have permission to use every uploaded face and will use the result responsibly.

What can you control in a multiple face swap?

Detect the people in one group photo, define each source-to-target relationship, and verify the complete map before generation.

Up to 5 Faces Detection

Detect and manage up to five visible faces in one group photo. Each target receives a stable number and cropped preview so you can select people without guessing their position.

Three Unique Mapping Modes

Replace one person, assign different source portraits to selected targets, or apply one source face across everyone you select.

Real-Time Relationship Mapping

The interface shows every source-to-target relationship before generation. Stable face numbers make it clear which uploaded portrait belongs to each detected person.

Upload with Confidence

Use clear JPG, PNG, or WebP group and replacement images. Confirm that you own the media or have permission from the people shown before detection begins.

Server-Verified Results

The server validates the detection session, face-number boundaries, and required source files before generation. Unselected faces remain unchanged in the submitted mapping.

Private and Secure Processing

Detection previews use short-lived signed links instead of public object URLs, keeping uploaded media within the private processing workflow.

Detection before generation

Every target face gets a stable number

The page does not ask you to guess positions such as “the person on the left.” Replicate returns one destination-face preview for each index from 0 through 4, and each source is mapped to one exact preview.

Numbering is an interface explanation, not a generated result. A real provider before/after remains a release gate.

Three mapping modes, one explicit relationship

Replace one person

Select one detected number, upload one source face, and leave every other person unchanged.

Different face per person

Assign separate source files to selected targets and verify every pair before generation.

One face across the group

Upload once, choose replace all, and let the server expand the source into all detected targets.

Relationship check

Read every mapping as source → target

Source A

Face 1

Replace

Source B

Face 3

Replace

No source

Face 2

Keep original

Real Replicate result

One selected person changed. The other two stayed in place.

This is a real indexed face-swap run from the tool: Face 2 was selected, one source portrait was uploaded, and the final JPEG was copied from Replicate into private result storage. No mock result is used.

Three people in the original group photo before a multiple face swap

Real Replicate multiple face swap result with one selected person replaced

How it works

Detect first, map second, generate last

Upload one group photo

Detection runs before credits are charged for generation.

Review faces and sources

Select targets, attach source portraits, and check the numbered relationships.

Generate selected swaps

Only mapped targets are sent; unmapped people remain unchanged.

Check the group before mapping

Five clear faces are more reliable than a crowded image with tiny, hidden subjects.

All intended faces are visible and not cropped

Each source portrait has one clear face

Source and target angles are reasonably similar

Every selected face number has the intended source

Useful when the relationships matter

Consenting group edits

Replace selected people without changing everyone in the photo.

One identity across a cast

Apply one authorized source face to every detected target for a deliberate visual concept.

Character assignments

Give different people different replacement faces and preserve the source-to-target plan.

The server verifies the map

The client cannot invent provider URLs, detection IDs, or out-of-range face numbers.

Detection must belong to the same requester

Target index must exist in the detected set

Each source key must match an uploaded file

A request can map at most five target faces

Private inputs

Detection previews use short-lived signed links rather than public object URLs.

2 credits

Credits are reserved atomically at generation and refunded once on a clear provider failure.

Shared guest trial

The AI trial is shared across face-swap pages rather than reset for each tool.

Replicate pipeline

Detection and indexed swaps use one provider

Replicate detects the numbered faces first, then applies each selected replacement to its destination index in order. Unselected people remain unchanged, and no client-supplied provider URL is trusted.

Replicate only

Use a narrower workflow

Choose single-image or animated face swap when a group map is unnecessary

01Photo Face SwapUse a simpler target-plus-source workflow for one face.02GIF Face SwapReplace one face across an animated GIF or WebP.03Batch GIF Face SwapApply one face to several animation files in a batch.

Questions before you start

Multiple face swap FAQ

Numbering, mapping, limits, and consent are the key decisions before group generation.