GPT Image 2.5 is live — OpenAI's newest image model, targeted edits that leave the rest of the frame alone
All resources

· reAPI Team

AI Face Swap: How It Works, Free Tools, API Price and the Rules

AI face swap and head swap explained: open-source tools and their licenses, a $0.013-per-image API, tips for clean results, and the EU rule on deep fakes.

AI Face Swap: How It Works, Free Tools, API Price and the Rules

An AI face swap takes the face from one photo and places it onto a person in another, keeping the second photo's pose, lighting and background. A head swap goes further and replaces the whole head, hair and outline included.[1][2] You can run one with free open-source software, in a consumer app, or through an API for about a cent per image.

The technology is the easy part. The license of the model you use and the rules about showing swapped faces to other people are where projects get into trouble, and both changed recently: the EU's disclosure obligation for deep fakes has applied since August 2, 2026.[3] This guide covers the tools, the cost and the rules, as checked on November 4, 2026.

Face swap vs head swap

The two differ in what they carry across:

Face swapHead swap
What movesThe faceThe whole head, including hair and outline
What stays from the target photoPose, framing, lighting, hair, backgroundBody, pose, lighting, background
Use it whenThe person's face is what mattersHairstyle is part of the identity, such as a recurring character

Based on the descriptions of reAPI's two models.[1][2] Moving a face onto a target with very different hair can look wrong even when the face itself is convincing; that is the case head swap exists for.

Free AI face swap tools and their licenses

The best-known open-source option is FaceFusion, which describes itself as an "industry leading face manipulation platform" and had over 30,000 GitHub stars and recent commits in November 2026. Its repository is licensed under OpenRAIL-AS, so read those terms before building on the software.[4]

Check the models as well as the software. InsightFace, a widely used face-analysis project, releases its code under the MIT license, but states that models trained on its data are "available for non-commercial research purposes only", including models downloaded automatically by its Python library.[5] A free tool that pulls such a model in the background is not automatically free for commercial use.

Free web apps that promise a face swap with no sign-up are another route. Before uploading anyone's photo, find out who runs the site, how long it keeps uploads and whether it trains on them; a face is personal data in many jurisdictions.

AI face swap by API

For an app or a batch job, an API avoids running GPUs. reAPI's two models each cost $0.013 per generated image, so 1,000 swaps come to $13.[1][2] A request takes two public image URLs and an optional seed:[6]

curl https://reapi.ai/api/v1/images/generations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"model": "face-swap", "image_url": "https://example.com/target-photo.jpg", "face_image_url": "https://example.com/source-face.jpg"}'

The documented limits are worth knowing before you design around it. It works on still images only, returns exactly one image per request, has no prompt or size control, and rejects base64 and data: URIs. Results are stored in an isolated bucket and their URLs last 30 days, so copy files you want to keep. Polling for the result does not use credits.[6]

Getting a clean result

The source face matters most. reAPI's documentation recommends "a clear, front-facing shot in even light" for the strongest result.[6] Beyond that, our suggestions:

  • Match the angle. A face photographed head-on placed onto a profile pose has to be guessed in 3D and usually shows it.
  • Match the light direction where you can, since the target's lighting is kept.
  • Avoid hands, hair or glasses across the face in either photo.
  • Pin a seed once you get a result you like; the same two images and seed reproduce it.[6]

The EU AI Act defines a deep fake as AI-generated or manipulated image, audio or video content "that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful". Article 50(4) requires deployers of an AI system that generates such content to disclose that it has been artificially generated or manipulated. For evidently artistic, creative, satirical or fictional work, the obligation is limited to disclosing the existence of such content "in an appropriate manner that does not hamper the display or enjoyment of the work".[3] The regulation applies from August 2, 2026.[3]

Disclosure is the floor, not the whole job, and the Act places it on the deployer, the party using the system to make the content. Get permission from the person whose face you use, keep a record of it, and label results when you publish them. reAPI's own documentation states the same expectation: use the model "on people whose likeness you have the right to use".[6]

FAQ

It depends on where you publish and what you do with the result. In the EU, deployers must disclose that deep-fake content is artificially generated, with lighter requirements for clearly creative or satirical work.[3] Using someone's face without their permission can also breach privacy, publicity or data-protection laws, so get consent first.

Can I face swap a video?

Not with reAPI's model, which handles still images only.[6] Some open-source tools, such as FaceFusion, are built for broader face manipulation; check their documentation for video support.

How much does an AI face swap API cost?

On reAPI, $0.013 per output image for either Face Swap or Head Swap, with one image per request and no charge for polling.[1][6] Running an open-source tool on your own GPU has no per-image fee, but you pay for the hardware.

Which is better, face swap or head swap?

Neither in general. Use face swap when the face is what matters and the target's hair suits it; use head swap when hairstyle is part of who the person is.[1][2]

Choosing an AI face swap route

For experiments on your own photos, an open-source tool is free, once you have checked its license and the license of the models it downloads. For a product, an API keeps the cost per image predictable: Face Swap and Head Swap on reAPI are $0.013 each. Whichever you use, an AI face swap that other people will see needs the subject's consent and, in the EU, a clear label.

References

  1. reAPI. Face Swap model page. Retrieved November 4, 2026 from reapi.ai/models/face-swap
  2. reAPI. Head Swap model page. Retrieved November 4, 2026 from reapi.ai/models/head-swap
  3. European Union. Regulation (EU) 2024/1689 (Artificial Intelligence Act), Article 3(60), Article 50(4) and Article 113. Retrieved November 4, 2026 from eur-lex.europa.eu/eli/reg/2024/1689/oj
  4. FaceFusion. facefusion/facefusion repository and license. Retrieved November 4, 2026 from github.com/facefusion/facefusion
  5. InsightFace. deepinsight/insightface repository README (license). Retrieved November 4, 2026 from github.com/deepinsight/insightface
  6. reAPI. Face Swap API reference. Retrieved November 4, 2026 from reapi.ai/docs/face-swap
All resources