AI Video Swap: How to Turn Into Anyone Holding Any Product

I recorded a few seconds of myself holding a book and got back a video of a different woman holding the same book and moving exactly the way I did. The tool behind it is Video Swap inside the AI Media Machine. You give it a reference video and the character you want to turn into, click swap, and it rebuilds the clip with the new person.

The short version of the result is below. The full tutorial is the video at the top of this page.

What you need before you start

You need a driving video and a snapshot taken from it.

Record the driving video on your phone or on a camera. I made two recordings. In the first I'm just talking with nothing in my hands. In the second I'm holding the thing I want to advertise. For me that was a book, but any product works.

Then take a snapshot of each recording, so you have one still of yourself talking and one of yourself showing the product.

It doesn't matter what you're wearing or what the room looks like, because the AI swaps all of that out. You tell it to replace the character and keep the product.

The driving video is also how you control the motion. Say you want the character to stop, say "oh, I forgot this," and point at your product, or open the product in a certain way, depending on what you sell. You record yourself doing that action and the AI copies it. In my results, the women on screen perform the exact motion I performed.

Step 1: take the snapshots

I recorded my clip on my phone. To grab a still on a Mac, press Command + Shift + 4 and drag over the frame you want. On Windows, use whatever screenshot shortcut you normally use.

Step 2: pick or create a character

If you already have an image of the person you want to turn into, use that. If not, open the AI Avatar tool. It has plenty of realistic-looking people to choose from, and you can generate new ones quickly. I created a custom character with this prompt:

UGC selfie 25-year-old influencer woman

If one of the existing avatars fits, right-click the image and save it. Either way, download the character so you can upload it in the next step.

Step 3: swap yourself out in Image Maker

In Image Maker, upload your snapshot, then upload the character as the reference image. My instruction was:

Replace the man with the woman from the reference image. Keep the man's background the same.

I chose 16:9, and 1K resolution was enough. The result looked very realistic, and she was even wearing my clothes.

For the product shot, I deleted my original snapshot, uploaded the one where I'm holding the book, and left the other settings almost the same. I added one sentence to the prompt:

Keep the book in the hand of the woman.

The AI kept the book in her hand. She wasn't wearing my clothes this time, since I gave it a different snapshot than the first image. That was fine, because the next step only uses the shot with the book.

Step 4: run Video Swap

Open Video Swap and upload your driving video first. This is the clip that drives the motion. Video Swap is meant for short clips of around 4 or 5 seconds, maybe 10.

Next, upload the character image. Everything you create is saved in your Library, so you can download it from there if you need to.

There's a prompt field, but I usually leave it as it is. Then set the character orientation to follow video or follow image. I chose follow video because I wanted my exact motion to stay intact. I also left "keep original sound" checked, then clicked generate.

Rendering takes up to 2 minutes for every second of video, so expect a short wait.

While it renders: Video Maker

Video Maker is a separate tool with three quality levels. Fast is the least expensive way to make a video. To test it, I added the character image and gave her a line in quotes:

"The AI Media Machine is the best AI video generator tool."

I set the length to 6 seconds, left audio generation on, and clicked generate. It takes about 20 seconds, sometimes a little more, and the result already looked good. If you don't need complex motion, Fast is enough.

Ultra is the highest-quality level, and it's the one to use when you want the most realism. Write the dialogue in quotes, add one frame of the character, and choose the duration. I recommend shorter clips of 5 to 6 seconds. Ultra takes longer to render. The blonde version in the video came from Ultra, and she says, "I think AI Media Machine is the first AI video ad tool where your ad doesn't look fake."

The result, and one problem with the voice

When the swap finished, she moved exactly the way I moved in my driving video. That's why Video Swap works well for ads: you decide the motion and the AI follows it.

There was one problem. Since I kept the original sound, she was speaking with my voice, which obviously doesn't fit her face.

Voice Changer fixes that. Upload the video and the voice you want, and it replaces the audio.

If you want the same voice across all your videos, you can upload your own voice sample. Record at least 10 seconds of audio, upload it, choose the voice, and click switch voice. The tool keeps that voice consistent and syncs it to the lips.

Try it

The AI Media Machine is available as a lifetime deal at 94% off. You pay once and there's no subscription. If you'd rather have my team produce the videos for you, book a free strategy call.