How to Make an AI Face Swap Video from Two Photos

A convincing face swap should change one thing—the facial identity—while keeping the target photo recognizably the same. The hairstyle, expression, head angle, body, pose, clothing, camera, light and background should still come from the target. That preservation goal is what separates a controlled two-photo edit from a generator that redesigns the whole portrait.

This guide answers the tutorial intent behind “how to make an AI face swap video.” It focuses on choosing the two inputs, assigning them correctly and checking both the replacement face and the unchanged scene.

Target photo versus source face

The two inputs do different jobs. The target is the scene you want to keep. The source is the facial identity you want to place into that scene. Reversing them produces the opposite result, so label the intended final image before uploading anything.

On LumiChat, both slots can use an authorized upload or an eligible realistic character from the character library. The first release expects one clearly visible adult in each image, not a group photo.

Choose a target photo worth preserving

Pick the target for its composition, not for the face you plan to replace. Use a scene whose pose, clothing, crop and background you already like. The face should be large enough to inspect, with the full head and hairline visible. Heavy motion blur, a tiny face, extreme shadow or a hand covering the jaw makes preservation harder.

The output follows the target’s square, portrait or landscape orientation. Do not crop the target only to make its face match the source; choose a source that fits the target instead.

Choose a compatible source face

The source should show the facial identity clearly. Similar head direction and camera height are more useful than matching clothing or background, because those source details are intentionally ignored. A front-facing, evenly lit adult portrait is the most flexible starting point.

Avoid sunglasses, masks, hair across the eyes, beauty filters that reshape the jaw and low-resolution social thumbnails. An extreme side profile can work only when the target has a compatible profile. Source expression does not need to match exactly because the target expression remains the goal, but a relaxed mouth and visible jaw give cleaner evidence.

How to make an AI face swap video step by step

1. Open the LumiChat AI Face Swap.
2. Place the image whose scene you want to keep in Target photo.
3. Place the authorized adult identity you want to transfer in Source face.
4. Confirm that both inputs show one realistic adult and that you have permission to use them.
5. Choose Image for a still result or Video for a five-second animated result.
6. Review the displayed credit estimate, accept the required use policy and start generation.
7. Check identity and scene preservation before downloading or sharing.

The best practical rule is match the source to the target, not the target to the source. The target already contains the valuable pose, wardrobe and scene. Changing it to accommodate a poor source removes the very context the swap is supposed to preserve.

Image versus five-second face swap video

Image mode creates one face-swapped frame. Video mode first creates that frame internally and then uses it as the starting point for subtle breathing, a blink and a small head movement in one continuous five-second clip. Only the final video is shown for the video task.

This is not uploaded-video face replacement. You do not provide a source video, and the tool does not preserve original motion, timing or audio. Choose Image for portraits, avatar concepts and fast quality checks. Choose Video only after the still composition is convincing and you want a small amount of natural motion.

Review identity and preservation separately

Do not judge the result with one vague “looks good” test. First compare the new face with the source: face shape, eye spacing, brows, nose, lips and jaw. Then cover the face and compare everything else with the target: hair, expression, head angle, pose, clothing, crop, objects, shadows and background.

Finally, watch the video for identity drift, hair movement, wardrobe changes, background pulsing or camera cuts. A result can have a similar face but fail preservation, or preserve the scene but fail to transfer identity. Both checks must pass.

Common face swap problems and fixes

Change one input at a time. If the source similarity is weak, keep the target and replace only the source. If the scene changes, keep the source and choose a stronger target.

Creative uses, consent and disclosure

Authorized face swap works for original-character casting, private portrait concepts, costume previews, avatar exploration and short fictional clips. It should not be used as evidence that a real person wore an outfit, attended an event or performed an action. If the real goal is to change clothing while keeping the same identity, follow the AI clothes changing guide instead of swapping faces.

Use only clearly adult, realistic people and content you own or are authorized to process. Obtain the depicted adults’ consent for the edit and any sharing. Do not use minors, age-ambiguous people, recognizable public figures or unsuspecting private individuals. Never use face swap for impersonation, fraud, harassment, blackmail or misleading identity claims. Review the current Privacy Policy and disclose AI editing when viewers could mistake the result for a real photo or event.

Current pricing can change, so use the estimate inside the tool rather than a number copied from an old guide. The pricing page explains wider account options, while the generator remains the source of truth for the selected task.

FAQ

Which photo is the target in a face swap?

The target is the photo whose hair, expression, body, pose, clothing, framing, lighting and background you want to keep.

What does the source face provide?

It supplies facial identity cues such as face shape, eyes, brows, nose, lips, cheekbones and jaw—not hair, clothing, pose or scene.

Can I make a face swap video from two photos?

Yes. LumiChat first creates the swapped image internally and then animates it into a continuous five-second clip.

Can I upload an existing video for face replacement?

No. The current workflow uses two image references and does not preserve the motion or audio of an uploaded source video.

Can I swap one face in a group photo?

Not in the first release. Each input must show exactly one clearly visible adult.

Why did the hairstyle change?

The source hair may overlap the facial boundary or the target hairline may be unclear. Use a source with the face unobstructed and a target with the full hairline visible.

Can I use a LumiChat character as either input?

Yes, when the original character is realistic and eligible for this tool. Authorized uploads are also supported in both slots.

Do I need permission to use someone’s face?

Yes. Use only authorized adult images, obtain consent for generation and sharing, and never use the result for impersonation or deception.