How to pick the right video and photo for an AI remake

An AI remake model does two jobs at once. It reads how the person in the original video moves, and it redraws that motion as you, based on your photo. Anything that makes either job harder shows up in the result. The good news is that a few simple rules avoid almost every bad result.

Choosing the video

One person

The models behind SynthGen AI follow one body through the clip. With two or more people in the frame, the AI has to decide whom to follow, and it may switch halfway. Solo clips are the safe choice. Busy backgrounds are fine, as long as one person is clearly the subject.

The whole body in view

Head to feet is ideal. When hands, feet or the head leave the frame, the AI has nothing to read for that moment and has to invent it, which is where blur and odd limbs come from. Clips filmed from a little distance work better than tight close-ups for dances.

One continuous shot

This is the most common mistake. Many trend videos cut between angles or between people. The AI follows motion through one continuous shot, so at a hard cut the motion stops. Scrub through the clip before you share it, and pick one filmed in a single take.

Clear, readable motion

Moves the camera can see clearly come out best. Very fast spins, heavy motion blur, strobe lighting and dark rooms make the pose harder to read. A steady camera helps too: a slow pan is fine, a shaky phone whipping around is not.

3 to 30 seconds

SynthGen AI takes clips from 3 to 30 seconds and uses the first 30 seconds of anything longer. Shorter clips finish faster and cost fewer credits, because the cost is based on length. If the best moment is in the middle of a long video you have saved, trim it in Photos first and pick it with From Photos.

Choosing your photo

Sharp and front facing

Your face is the hardest part to get right and the first thing people notice. Use a photo where you look roughly toward the camera, in focus, with your eyes visible. A slight angle is fine, a profile is not.

Even light

Soft daylight from a window is perfect. Avoid hard shadows across the face, strong colored lights and heavy beauty filters. The AI copies what it sees, including a filter’s smoothing.

Nothing covering your face

Sunglasses, a hand, hair over the eyes or a mask all hide information the AI needs. It will guess, and guesses look less like you.

Match the photo to the mode

• Swap me in rebuilds you inside the original scene. A clear portrait or upper body shot works well.
• Copy the moves uses your photo as the first frame, so what is in the photo is what moves. A full-body photo, standing, with your arms visible and away from your body, gives the cleanest animation. Your outfit and background will be in the video, so pick ones you like.

A 20 second checklist

Before you tap Make it me:

1. Is there one person in the clip?
2. Is their whole body in view most of the time?
3. Is it one shot, with no cuts?
4. Is it between 3 and 30 seconds, or is the good part at the start?
5. Is your face sharp, lit and uncovered in your photo?

Five yeses and you are very likely to love the result. If something goes wrong anyway, a failed video returns its credits automatically. Ready to try? See how to put yourself in any TikTok or Instagram video.