Skip to main content

Video Face Swap

Commercial use OK400+ modelsNo watermarkNo sign-up needed
Consent first — only swap faces of people who agreed to it. Deepfakes of real people, non-consensual content and swaps of public figures are refused. Working from a photo rather than a clip? Our photo face swap is live today.
Upload a clip of 4–15 seconds and one portrait. The AI re-renders the clip with the new face, following the original head movement and expression. It works generatively, so hair and lighting around the face can shift too — made for consensual creative edits, film VFX and fictional characters.

Drag a video here or click

Click here to choose a video

MP4 / MOV / WebM · 4–15s · max 100MB · one clear face works best

Drag a portrait here or click

Click here to choose a portrait

JPG / PNG / WebP · max 10MB · clear forward-facing shot

That means the person whose face is in the photo agreed to this. Every request is logged with the attestation. Don't tick this for public figures, celebrities or anyone who hasn't agreed — those are refused anyway.

Upload both files and tick consent to continue.

Responsible uses for AI face swap

Stunt doubles and VFX

Put the lead actor's face on the double in post. Rotoscoping the same shot by hand runs into days of work; a draft cut comes back in minutes, with everyone on set having signed for it.

Parody your own friends

Drop yourself or a willing friend into a scene from a film or a music video. Keep it obviously fictional, and say it's AI when you post it — that's the difference between a joke and a problem.

Recasting and test reels

See how a different actor reads in a scene before anyone books a reshoot. Useful for storyboards, casting conversations and a director's own what-if cuts.

How to swap a face in a video

1
Uploading the target video

Drag and drop the clip of the person whose face you want to swap — MP4, MOV, or WebM, up to 100 MB and between 4 and 15 seconds long. The most accurate face swap is when the face is clearly visible and roughly facing the camera.

2
Uploading a new face

Drag and drop one portrait of the person you want to appear in the clip: full-face, well-lit, without sunglasses or strong shadows. JPG, PNG, or WebP files up to 10 MB in size.

3
Confirm consent and press "Swap faces"

Tick the box to confirm the person's consent, then press "Swap faces". The finished clip comes back without a watermark, and the original audio track is carried through untouched.

Why not one of the apps or the open-source scripts?

The phone apps are the quick option, and most of them either stamp the result with a watermark or put the clean export behind a subscription. The open-source projects have no such catch and give you the most control, but you own the setup: Python, CUDA, model weights and a GPU that can sit there for the length of the render. This one runs in the browser with nothing to install, gives you the clip without a watermark, costs tokens from the free daily pool rather than a monthly plan, and asks who consented before it starts rather than after someone complains. Working from a still photo rather than a clip? Our free AI face swap does that one today.

Use this tool via API

Call this tool straight from your own scripts. The endpoint follows the OpenAI REST format and signs requests with a bearer token, so client libraries you already use keep working. You are billed the same token rate as the web app.

API DocumentationGet API Key
curl
curl -X POST https://api.rewind.ai/v1/video/face-swap \
  -H "Authorization: Bearer sk-rewind-..." \
  -F [email protected] -F [email protected] -F consent=true

Related Free AI Tools

AI Video Generator
AI Video Editor
Image to Video
AI Video Dubbing Studio
AI Video Avatar
AI Video Stabilizer
AI Video Effects — Pikaffects-style
AI Video Meme Generator
See all video tools

520+ AI Models

Chat, image, audio and video models, all reachable through one API.

Qwen 2.5Chat, 7B-72B
FLUXImage gen
CogVideoXVideo gen
KokoroTTS, 54 voices
WhisperSpeech-to-text
Mistral LargeChat
DeepSeek R1Reasoning
Lyria 3Music gen

Plus 520+ models via external providers: GPT-6, Claude, Gemini, Llama, and more.

Explore all models

Simple Pricing

Start free. Pay only when you need more.

Free5K tokens/daySign up free
Basic$5/mo, 200K tokens/moGet started
Most PopularPro$19/mo, 1M tokens/moGet started
Business$49/mo, 5M tokens/moGet started
View all plans

FAQ

You give it a short clip and one portrait, and the model re-renders the clip with the new face in place — following the head as it turns and keeping the original expression and timing. It's the video counterpart of the photo face swap we already run, though it works differently under the hood: the photo tool edits one still, while this one regenerates the footage.

Yes, and it's a hard gate rather than a note in the terms: the button stays disabled until you tick the attestation, and every request is logged with it. Only swap the face of someone who agreed to it.

No. Swaps involving public figures are refused whether or not the box is ticked, as is anything sexual or built to pass as real footage of a real person. For a face nobody has to consent to, generate a fictional one with our image tools instead.

Yes, on the same terms as the rest of the site: a daily token pool that resets every 24 hours, no card and no trial. Signing in for free triples the pool, and paid plans start at $5/month if you need more.

MP4, MOV and WebM, up to 100MB, and between 4 and 15 seconds long. That window is the model's, not a policy of ours. The reference face is a JPG, PNG or WebP up to 10MB. If your clip is longer, trim it to the section you actually need before uploading.

It depends mostly on your inputs. A clip with one well-lit face looking roughly at the camera, and a sharp forward-facing portrait, give the cleanest result. Heavy motion blur, faces in profile, sunglasses and partial occlusions are where any face-swap model starts to show seams.

No — it works on one face, the most prominent one in the clip. Be aware that the model re-renders the whole clip rather than patching a face-shaped area, so other people and the background can shift slightly even though only the one face is being replaced. Check the finished clip before you use it.

Fast pans and quick head turns are the hardest case, because there's less of the face in each frame to work with. Steadier footage gives a noticeably better result, so trim to the calmer section if you can.

Yes. Only the picture is re-rendered; the original audio track is carried across untouched. The one exception is a WebM whose audio can't travel in an MP4 — that gets re-encoded to AAC, and the result says so. To change what's said, that's a different tool — see the AI Video Dubbing Studio.

Your files are processed on our own GPU servers and aren't kept permanently. We don't sell them and we don't train models on them. The consent attestation itself is logged with the request.

A developer API is in progress across the site, using the same Bearer-token REST endpoints as the other tools. Get in touch if you want early access.

Love Rewind.ai? Tell your friends!

Rate this page