How to Face Swap a Photo Online: Complete Guide (2026)
A practical photo face swap guide covering source and target preparation, batches up to 20 images, model choice, face mapping, and review.
Photo face swap workflow
Correct target-to-source mapping and a full-size review matter more than a thumbnail preview.
- 1Upload an authorized target photo with clearly visible faces.
- 2Add clear, authorized adult source portraits.
- 3Map every detected target face to the intended source and confirm the model and cost.
- 4Generate, then compare hairlines, eyewear, skin tone, lighting and proportions at full size.
- 5Adjust the source or mapping and retry only the photos that need it.
Photo face swap uses a source portrait to replace a detected face while attempting to match the target pose, light, and expression. Framedance processes up to 20 target images per batch, offers realistic and high-similarity styles, and supports single- or multi-face mapping. Those style names describe optimization goals, not guaranteed results; profiles, occlusion, small faces, and strong lighting differences still need image-by-image review.
TL;DR
Swap faces in up to 20 photos in a single batch run — one setup, twenty results.
Two model tiers: realistic for seamless blending, and high-similarity (the pro option) when the output must look like the actual person.
Multi-face mode maps every face in a group photo individually, and cross-image dedup means each person is mapped only once.
Recently used source faces are saved for one-click reuse — recurring projects get faster every time.
The workflow provides model-style choices, batching, and downloadable results; all inputs and uses remain subject to the content rules.
How do you face swap a photo online?
- Open Framedance's image face swap tool and sign in.
- Upload your target photos — up to 20 images in one batch.
- Choose single-face mode for one swap per photo, or multi-face mode to map each detected face separately.
- Upload a source face photo, or pick one of your recently used faces for one-click reuse.
- If the same person appears in several photos, map them once — cross-image deduplication handles the rest.
- Pick a model: realistic for natural blending, or high-similarity when likeness to the source person is the priority.
- Run the batch, review results in the on-page history, download what you like, and retry any failed image individually.
Realistic vs high-similarity: which model should you pick?
The realistic model prioritizes how well the new face blends into the target photo — skin tones, lighting, and grain all match, which makes it the safer default for stylized shots, artistic edits, and memes.
The high-similarity model is the professional option: it pushes the output to resemble the source person as closely as possible. Choose it for personal avatars, portraits, and client work where someone will judge the result by how much it looks like them.
A practical approach is to evaluate the available styles on one representative target, compare blend edges, identity resemblance, and lighting in history, then use the better fit for the remaining batch. Review the current charge before each submission.
How does batch and multi-face swapping actually work?
Batch mode accepts up to 20 images and treats each one as its own job, so one bad photo never sinks the run — you can retry failed items individually. Results accumulate in the history strip on the page, ready to compare and download.
In multi-face mode, every detected face in every photo gets a mapping slot. The clever part is cross-image deduplication: Framedance recognizes when the same person appears across multiple photos and asks you to map them exactly once. For an event album with five people across twenty shots, that means five mappings instead of dozens.
Recently used source faces remain available for reuse in recurring work such as a weekly series or product set. Before reuse, confirm that the selected identity and asset authorization are correct.
How much does photo face swap cost?
Photo face swap consumes credits according to the current model style and task count. Check the displayed estimate and validate the mapping on one representative image before releasing a batch.
Each image is an independent job, and cost varies with job count and current model. A failed item can be retried separately, but a retry may consume credits again, so first classify the cause as source quality, mapping, parameters, or a temporary service problem.
How do you get a convincing photo face swap?
Start with the source photo: front-facing, sharp, evenly lit, and with only one face in the frame. A selfie can work, but a well-shot portrait almost always swaps better.
Angles matter: a target closer to the source angle is usually easier to blend, while an extreme profile is harder. If several authorized portraits of the same person exist, choose the closest angle and reuse it through recent faces.
Finally, model choice and retries: switch to high-similarity when likeness matters, and if the face looks pasted on, the realistic model usually fixes it. Retry with another source photo before blaming the tool — two photos of the same person can behave very differently.
Can you run photo face swap through the API?
Yes. Create an API key on /api-keys, inspect the current model schema, validate one authorized request, and then integrate the model into your server-side application or script.
curl -X POST "https://www.framedance.ai/api/v1/marketplace/run/<model_id>" \
-H "Authorization: Bearer YOUR_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"input": {
"image_url": "https://example.com/target.jpg",
"face_image_url": "https://example.com/face.jpg"
}
}'
# Poll until the task completes
curl "https://www.framedance.ai/api/v1/marketplace/run/tasks/<task_uuid>" \
-H "Authorization: Bearer YOUR_API_KEY"Practice: map a small authorized photo set
Prepare three authorized target photos of adults: a clear frontal portrait, a three-quarter view, and a two-person image with both faces unobstructed. For each person, choose one sharp adult source portrait with even light and a neutral expression. Keep the source face large in frame and avoid filters, motion blur, cropped chins, covered eyes, and mixed lighting. This set tests pose and multi-face mapping without hiding the cause of a failure.
Upload the targets, inspect every detected face, and map each one explicitly to the intended source. In the two-person photo, confirm the left and right people are not reversed. Select the available model according to the task rather than assuming every model behaves the same, review the displayed price, and generate only this small set first. Do not use faces of anyone under 18 or a likeness you are not authorized to use.
Full-size acceptance checks and targeted fixes
Compare each result with its target at full size. Accept only when the correct identity is assigned, eyes and mouth align with the target pose, hairline and jaw transitions are unobtrusive, eyewear remains coherent, and skin tone, shadows, and proportions fit the surrounding photo. Check every person independently in a multi-face image. Identity consistency is not guaranteed, so approval must be based on the actual output.
A completely wrong person indicates a mapping error: fix the assignment. Distorted eyes or mouth usually call for a source with a closer angle and clearer landmarks. A visible skin-tone seam suggests more compatible lighting; broken glasses or hair edges may improve with a simpler, unobstructed source or target. Change one factor and retry one photo so you know what helped and avoid paying to repeat results that already passed.
Câu hỏi thường gặp
How is photo face swap billed?+
Cost depends on the current model style and task count. Use the credit estimate shown before submission and validate one representative image first.
Do the swapped photos have a watermark?+
Downloads are normally delivered without a Framedance watermark. Inspect the image and disclose synthetic media where platform rules or law require it.
Can I use face-swapped photos commercially?+
Yes, you can use them in commercial projects, as long as you hold the rights to both the target image and the source face. Always secure consent when a real person's likeness is involved.
What content rules apply to photo face swaps?+
Pornographic content, face swaps or sexualized content involving minors, and unauthorized use of real people are prohibited. Confirm that everyone is at least 18 and authorized, and do not use face swaps for fraud, impersonation, false endorsements, or identity-verification bypass.
How many photos can I swap at once?+
Up to 20 images per batch. Each image is processed independently and can be retried on its own, and with cross-image dedup plus one-click face reuse, large albums go surprisingly fast.