Group-photo workflow guide

How to swap multiple faces in one group photo

A multi-person edit succeeds when every replacement identity is mapped to the intended detected face. This guide turns that mapping into a visible, numbered workflow you can verify before spending credits.

By DeepSwapAI Product TeamReviewed July 14, 2026Practical guide
DeepSwapAI multiple face swap workspace showing a group photo and numbered face mapping controls
The live group-photo workspace detects faces first, then makes each replacement-to-target mapping visible before generation.

Make every intended target easy to identify

Start with the group photo, not the replacement portraits. If a target face is tiny, covered, blurred, or merged with another person, the mapping may be clear while the final identity evidence remains weak.

CheckPreferAvoidWhy it matters
Face visibilityEyes, nose, mouth, and jaw mostly visibleHands, masks, glare, or hair covering landmarksEach selected face needs enough structure for identity transfer.
SpacingFaces separated by visible background or clothingOverlapping heads or one face crossing anotherClear boundaries make detection and result review easier.
AngleFront or moderate turns with compatible referencesExtreme profiles paired with only frontal referencesLarge pose conflicts can weaken shape and identity cues.
ResolutionA photo where each intended face can be inspected at normal sizeUpscaled thumbnails or heavily compressed downloadsEnlarging the file does not recreate missing facial detail.
Unselected peopleA clear decision about who should remain unchangedSelecting every detection without checking its personOnly mapped targets should be replaced.

Map each replacement before you generate

  1. Upload one clear group photo. Use a JPEG, PNG, or WebP target where the people you intend to change are visible. The target image can be up to 30 MB.
  2. Run automatic face detection. Confirm that every intended person has a numbered detection. A missed or incorrect detection should be solved before adding replacements.
  3. Add one replacement for each selected person. Upload a clear identity reference for every target you plan to replace. The current workflow accepts up to 10 replacement images.
  4. Map replacements in the same numbered order. Select each detected target face in the same order as its replacement reference. Read the full mapping once from start to finish before submission.
  5. Review cost, generate, and inspect every face. The current rate is 2 credits per selected face. After generation, inspect the mapped identities, face edges, lighting, and every person who was supposed to remain unchanged.
Ordering rule: reference 1 must correspond to selected target 1, reference 2 to selected target 2, and so on. Do not infer the order from left-to-right position alone; use the visible numbered mapping.

Fix the failed decision instead of retrying blindly

The wrong person changed

Return to the numbered mapping and compare every replacement reference with its selected target. Rebuild the order explicitly before another submission.

An intended face was missed

Use a target where that face is larger, sharper, and less obstructed. Detection cannot reliably map a person it does not identify.

One identity looks weak

Replace only that person's reference with a sharper portrait whose pose and expression better match the target face.

Hair, glasses, or edges look wrong

Inspect whether the target hides facial landmarks or places moving objects over the face. Hair and accessories belong to the target scene and still need clean boundaries.

Unselected people look affected

Verify the selected detections and inspect the original at normal size. Keep the target count limited to the people you actually intend to replace.

The task did not complete

The current settlement is all-or-nothing: credits are charged only when every requested replacement succeeds. Read the live error, correct the cited input or mapping, and resubmit.

Know the complete group task before submission

ItemCurrent behaviorPractical check
Accepted imagesJPEG, PNG, and WebPUse the original file rather than a repeatedly compressed copy.
Per-image limit30 MBCheck the group photo and every replacement reference.
Total upload limit120 MBCount the target plus all replacement files.
Replacement limitUp to 10 identity referencesUse one clear reference for each selected target.
Current cost2 credits per selected faceThe amount displayed in the live workspace is the final pre-submission reference.
SettlementAll requested replacements must succeedCredits are charged only after the complete mapped result succeeds.

Check every person's permission and the exact export

Only upload faces you are permitted to use. DeepSwapAI does not use uploaded content to train AI models, and uploaded and generated media is removed from DeepSwapAI servers within 24 hours.

Trial image exports from an account with no completed credit purchase visibly contain DeepSwapAI.com and the account's full email address. The interface discloses that exact address before generation. After any completed one-time credit purchase, future image exports are watermark-free; existing trial exports are not changed retroactively. Review the privacy and consent checklist before sharing a group edit.

Checked against the live mapped workflow

The DeepSwapAI Product Team reviewed the current detection, numbered mapping, validation, queue, settlement, retention, and watermark behavior. The guide does not invent an accuracy score, completion time, or output guarantee. Product facts and limits were reviewed on July 14, 2026; see the verification methodology.

Map the whole group before you generate

Upload one representative group photo, confirm every numbered target, and review the exact cost and full-email trial watermark disclosure before submission.

Open multiple face swap