Skip to main content

Face Swap Pro API aligns facial landmarks in two photographs for a precise identity transfer, with optional enhancement disabled by default.

From $0.16/image

Workbench

Input

Image editing
1/1

Drop files or media links, paste, or browse · 1 file · Max 10 MB · JPG · PNG · WEBP

  • 1
1 / 1 files
For Pro, image_url supplies the photograph to edit, including the hairstyle, pose and setting you intend to keep. at least 1 characters
1/1

Drop files or media links, paste, or browse · 1 file · Max 10 MB · JPG · PNG · WEBP

  • 1
1 / 1 files
Choose a sharp reference_image_url for Pro with lighting close to the target; its face supplies the identity rather than a new composition. at least 1 characters
Enhance face
face_enhance is false by default on Pro; enable it only when you want an additional face-enhancement pass instead of the default untreated finish.
Safety checkerOff

SpicyAPI does not run a safety checker on this model. Your prompt and media go to the model as written, and the model itself decides what comes back.

Checking your account before enabling submission and pricing.

Output

Ready for a new run
wet-skin-closeupExample result

Related models and next tasks

Missing the model you need?

Two fields, one transfer, nothing applied afterwards

Pro is the smallest surface of any swap endpoint in this catalogue: a target image and a reference face, and there is no third field. No prompt, no seed, no format, and no retouch pass over the result. That is the product, not a gap in it — it is the tier to pick when your own pipeline does the finishing and an extra beautification pass would be something to undo.

Open the workbench
Two fields, one transfer, nothing applied afterwards
Endpoint
One edit call, asynchronous, one image back
Optional fields
None — the schema has exactly two properties
Billing
Per returned image, charged on completion
Content filtering
No filter added on our side
Two fields total
The whole request bodyimage_url and reference_image_url, both accepting JPEG, PNG and WebP. additionalProperties is false, so nothing else is even accepted.
No retouch pass
The transfer is what you get backNothing is smoothed or re-graded after the swap. If a later stage of yours does colour work, there is nothing here to fight it.
Nothing to tune
No seed, no prompt, no formatThere are no parameters to sweep, so quality work goes entirely into choosing and framing the two input photos.

What the Face Swap Pro API deliberately leaves out

  • No post-transfer retouch

    Skin is not smoothed and the result is not re-graded. Pipelines that already run their own grade get a clean input instead of a processed one they have to work against.

  • No parameters to get wrong

    Two URLs and nothing else. There is no combination of fields that can be misconfigured, and no sweep to run before you trust the output.

  • A schema that will not drift under you

    additionalProperties is false over exactly two properties. Integration code written against it has almost no surface left to break.

  • Same accepted formats as the rest of the family

    JPEG, PNG and WebP on both inputs, as public URLs, uploaded file URIs or inline data URIs.

  • Asynchronous like every other task here

    Submit, then take the result from a callback or by polling the task. One task returns one image.

  • Uncensored, with the duty on the caller

    We add no content scan. Consent and likeness rights stay with whoever sends the images.

When Pro is the right tier

  • A finishing stage already exists

    You grade downstream

    A finishing stage already exists

    Colour, grain and skin work happen in your own pipeline. An endpoint that pre-smooths the result gives you something to undo before you can start.

  • Fewer fields, fewer failure modes

    Reproducible integrations

    Fewer fields, fewer failure modes

    Two required properties and nothing optional means the request shape cannot drift between services or between environments.

  • You are measuring the transfer itself

    Comparative evaluation

    You are measuring the transfer itself

    Without a retouch pass in the way, what you are judging is the identity transfer rather than the cosmetic layer on top of it.

  • Consenting adults, no platform scan

    Creator and uncensored workflows

    Consenting adults, no platform scan

    Nothing is filtered on our side. Authorisation is a question to settle before the images are sent.

Calling Face Swap Pro from a backend

  1. Put the effort into the two photos

    With no parameters, input quality is the only lever. A frontal, evenly lit reference at adequate resolution moves the result more than anything you could have configured.

  2. Get both images addressable

    Public HTTPS URLs, the platform upload flow, or inline data URIs all work. You do not need to host the files yourself.

  3. Send an Idempotency-Key

    A network retry without one is how a single image gets billed twice. With one, the repeat returns the original task.

  4. Prefer callBackUrl over polling

    The task comes to you when it finishes, which removes both the scheduler and the polling latency.

  5. Separate bad inputs from transient faults

    No detectable face is invalid_request — retrying it will fail identically. upstream_unavailable is worth a retry.

Worked examples

  • A swap that feeds a colour pipeline

    Take the returned image straight into your own grade. Because nothing was applied after the transfer, your LUT lands on the transfer rather than on someone else's finishing.

  • Evaluating tiers on your own photos

    Run the same pairs through Pro and Plus and count how many results you would actually ship. Samples in a gallery are chosen to look good; your own difficult pairs are the real test.

  • High-volume, uniform source material

    When the input photos come from one controlled setup, the automatic grading in Plus has little left to fix, and Pro's plain transfer is the simpler thing to operate.

Face Swap Pro questions

Where Pro sits

Pro and Plus cost the same per image and take the same two required inputs. Plus adds tone matching, a format field and a seed, and decides face-versus-head on its own. Pro applies none of that. The choice is not about quality tiers — it is about whether you want the finishing done here or downstream.

Models in this series3

Image models2