How to face swap a video.

The short version

To face-swap a video, you need three things: a clear front-facing photo of the face you want to use, a target video (either one you upload or a pre-loaded scene), and a face-swap tool that can run the swap for you. The most common path in 2026 is to use a face-swap app or bot — send the selfie, pick a scene, and the tool returns the swapped video in a couple of minutes. Standin is one such tool (it runs inside Telegram and is pay-per-clip). Reface, FacePlay, and Wombo are others (apps with different strengths).

The full step-by-step is below, with screenshots' worth of detail and a checklist for what makes a good selfie (which is the single biggest factor in the result looking good or uncanny).

What is face-swapping, actually?

Face-swapping is the process of taking one face from a source image and digitally placing it onto a target face in a video, while keeping the target's expressions, head movement, and lighting consistent. Modern AI face-swap tools do this with a generative model that has learned the relationship between facial geometry and appearance — it doesn't just "paste" the new face on, it re-renders the face frame-by-frame so it matches the target's expression and angle.

The output is a video that looks like the target person has the source face. Done well, it's cinematic. Done poorly, it falls into the "uncanny valley" — the face looks slightly off, expressions don't quite match, the lighting is wrong.

How the technology works (in plain English)

You don't need to know the math to make a good face-swap, but a 30-second mental model helps. Most AI face-swap tools use a combination of:

• Face detection and landmarking — finds the face in both the source image and every frame of the target video, and identifies the eyes, nose, mouth, and jawline.
• Face embedding — converts the source face into a numerical representation of its "identity" (the part that makes it look like you, not like anyone else).
• Generative reenactment — re-renders the target face with the source identity, frame by frame, matching the target's expression, head angle, and lighting.
• Blending and color matching — adjusts the new face so its skin tone, lighting direction, and color grading match the target video. This is the part that takes a swap from "uncanny" to "cinematic."

Standin (and most modern tools) handle all of this in a couple of minutes per clip. The output is a 9:16 MP4, ready to post anywhere.

Step-by-step: how to face-swap a video

The exact steps depend on the tool. Using Standin as the example (it runs inside Telegram):

1. Open @standinmebot in Telegram and tap Start. The first time, you'll see a welcome message and a consent prompt. Read it. Tap confirm.
2. Tap /render to begin a new render. The bot will show you the scene picker (a few pre-loaded scenes like action hero, rockstar, astronaut) or an "upload your own" option.
3. Send a clear selfie — the one key input. See the "what makes a good selfie" section below for what to do and what to avoid. The bot will accept the photo and start the swap.
4. Wait about 2-3 minutes. The bot sends progress updates while it works. Most renders come back in under 3 minutes.
5. Get the result in the chat. It's a 9:16 MP4. Tap and hold to forward to any other chat, or save to your camera roll to post elsewhere.
6. (Optional) Score it in your editor. Most creators drop the clip into CapCut, add trend audio, trim to the beat, and post. The bot ships silent by design (algorithm and licensing both work against baked-in audio).

What makes a good selfie for a face-swap

The single biggest factor in whether your swap looks cinematic or uncanny is the quality of the source selfie. Here's what matters:

• Front-facing, looking at the camera. The source pose has to match the target scene's pose closely. A 3/4 angle or a side profile will produce a worse result than a head-on, eyes-on-camera shot.
• Even, soft lighting. Harsh shadows on one side of the face make the swap look uncanny. Face a window, or use a ring light. Avoid direct overhead lighting (creates eye shadows) and direct under-lighting (creates the horror-movie look).
• No sunglasses, no hats, no hands blocking the face. The AI needs to see the full face. If something is occluded, the swap has to guess what to fill in, and the guess is often wrong.
• Neutral expression, or matching the scene's vibe. A slight smile is fine. Extreme joy or grimaces don't blend as cleanly. If the scene is "rockstar on stage," send a selfie where you look confident; if the scene is "shocked reaction," send a selfie where you actually look a bit shocked.
• Recent photo. Don't use a 5-year-old photo if you look meaningfully different now. The swap will look like the older you on the target's body, and the mismatch shows.
• Front-facing camera, not a group photo. Cropping yourself out of a group photo usually picks up other faces' geometry and the result looks weird. Take a fresh selfie.

What scenes work well

Not every scene is a good match for face-swapping. The best scenes share a few traits:

• The target face is front-facing or near-front-facing in the shot. The AI has to map your face onto the target's pose; a near-frontal source pose maps cleanly, a side profile or a face looking away doesn't.
• There's only one clear face in the frame. Multiple faces confuse the model — it doesn't always pick the right one.
• The lighting is consistent throughout the clip. A scene that goes from indoor to outdoor mid-clip will give the model a hard time color-matching.
• The clip is short (5-15 seconds). Most face-swap tools, Standin included, are optimized for short cinematic clips. Long videos take longer and the consistency drops.

Standin's pre-loaded scenes (action hero, rockstar, astronaut) all match these criteria. If you're sending your own video, look for one that does too.

For specific use cases:

• For TikTok creators — see the face-swap for TikTok creators guide.
• For wedding videos — see the wedding face-swap page.
• For birthday videos — see the birthday face-swap page.

Common mistakes and how to avoid them

Most bad face-swaps come from a small set of common mistakes. Here's what to watch for:

• Side-profile selfie. The AI has to rotate your face to match the target. The result looks like a poorly-pasted cutout. Fix: take a head-on selfie.
• Sunglasses, hats, or face coverings. The AI can't see the eyes, so it guesses. The guess is usually a flat black bar or a weird "mask" effect. Fix: take the sunglasses off, push the hat back, take the mask off.
• Harsh side lighting. The right side of the face is bright, the left is in shadow. The AI can't tell which is your "real" lighting, so the result has inconsistent shadows. Fix: face a window or use even lighting.
• Old photo, or heavily filtered photo. The swap looks like a different version of you. Fix: recent, lightly edited photo.
• Photo of a photo. The screenshot of a screenshot introduces compression artifacts that the AI can't always handle. Fix: send the original photo file, not a screenshot.
• Group photo cropped to just you. The model picks up other faces' geometry and the result has weird blending. Fix: take a fresh solo selfie.

After the swap: sharing, posting, scoring

Once you have the result, the question is what to do with it. Most users do one of three things:

1. Forward it to a group chat. Tap and hold, forward, done. The result is a real video file that works in any chat app (Telegram, WhatsApp, iMessage, Signal, Discord).
2. Save to camera roll and post to TikTok / Reels / Shorts. The output is vertical 9:16, so it drops straight into any of those platforms without a re-export. Add trend audio in the platform's editor, trim to the beat, post.
3. Use it as a personalized gift. Send it directly to the recipient (parent, partner, friend) with a personal message. This is the most common use for Father's Day, weddings, and birthdays.

If you want to use it for a professional use case (wedding photographer add-on, party planner activation, realtor listing), see the audience pages for specific workflows.

Is it legal? Is it safe?

Face-swapping is legal in most jurisdictions, but the use of the result is what determines legality. The general rule:

• You must own or have permission to use the source face (the selfie) and the target video (the scene).
• You cannot use the result for fraud, defamation, non-consensual intimate imagery, or to impersonate someone for endorsement.
• You cannot use the face of a minor in any context.

For the full safety story (what Standin does and doesn't allow, how the takedown process works, what to do if you're the subject of a render you didn't consent to), see the Trust & Safety page.

Frequently asked questions

How long does it take to face-swap a video?

About 2-3 minutes per render with most modern tools, including Standin. The tool sends progress updates while it works.

Is face-swapping free?

Most tools have a free tier. Standin's first render is free (with a watermark). Paid renders start at $9. Reface, FacePlay, and Wombo have free trials with daily swap caps.

Do I need to install anything?

Depends on the tool. Standin runs inside Telegram, so you don't need to install anything new. Reface, FacePlay, and Wombo are mobile apps you download. DeepSwap is a web app you open in a browser.

Can I do this on my phone?

Yes. Standin is built for phones (iOS and Android via Telegram). Most face-swap tools have a mobile-first design.

What's the best face-swap app in 2026?

Depends on the use case. For the biggest template library, Reface. For Telegram-native quick swaps, Standin. For the full ranked roundup, see the roundup page.