How to make an inserted photo look like someone placed it there by hand

Last updated 2026-09-15 · 8 min read

What is the short version?

An image that slides in at a steady speed and parks in the corner announces itself as an overlay. The same image brought in from the bottom left, held, and taken away again reads as a thing somebody showed the camera.

1. Go to Vizard Agent with the clip and the picture.
2. Say it should enter as though placed by hand.
3. Say how big it should be and when it leaves.

What do you need before you start?

The picture, the clip and two numbers: how much of the frame it should fill and when it should go. Vizard Agent works to those rather than to a feel, and both are the sort of thing that is obvious to you and invisible in a brief.

• The image. At a decent size.
• Which corner it comes from. Bottom left reads as a hand.
• How much of the frame it fills. A quarter, a half.
• When it leaves. A second, ideally.
• Whether it comes back. Once, or as a fixed badge.

What do you type into Vizard Agent?

Describe the movement as a person, not as an animation. "Slide the photo in" gets you a slide; saying it should look as though somebody is holding it into shot changes the motion, the timing and the angle together.

Prompt

Variants worth knowing:

• "As if a person were placing it." Changes the motion.
• "Fill the bottom half." A size you can check.
• "Confirm it is gone." The check that catches a leftover.

What does Vizard Agent actually do?

Here is the sequence Vizard Agent worked through on a real clip where a document had to appear in shot. Almost every pass came back to the same three things: how it entered, how much of the frame it filled, and when it left again.

1. Reviews the clip to see where the action and the edges are.
2. Opens the image to decide how it should be brought in.
3. Composites it with a hand-placed entry rather than a slide.
4. Extracts frames from the result to review.
5. Checks the entry and the settled position.
6. Rebuilds at the stated size when it lands too small.
7. Times the exit so it is gone by the moment you named.
8. Verifies on frames that nothing survives past that point.
9. Crops part of the image if it turns out to be too tall.
10. Re-checks the entry and exit timing after the crop.

Step seven is the half most people forget to ask for. An image that arrives well and then simply stays becomes furniture, and the exit is what keeps it feeling like a moment rather than a layer that was switched on.

Step eight is not paranoia. An overlay that should be gone at five seconds and lingers for two frames afterwards is exactly the kind of fault that survives a watch-through and shows up on a phone, so the frames get checked rather than the settings.

Step nine is worth knowing about. If the picture is the wrong shape for where it sits, cropping part of it is usually better than shrinking the whole thing — you keep the presence and lose the part nobody was looking at.

What does the result look like?

A picture that appears to be brought into the shot and taken away again. It fills the part of the frame you asked for, it has gone by the second you named, and nothing about it reads as an overlay that was keyed on.

Vizard Agent tells you the exact in and out frames, so shifting it by half a second is a precise instruction.

When does this not work well?

This is a look, not a simulation. There is no real hand and no real shadow, so on a shot where the light is strong and directional the missing shadow gives it away, and a static camera hides the trick better than a moving one.

• Strong directional light. A missing shadow shows.
• Moving camera. The image does not move with the room.
• Very busy frames. There is nowhere for it to sit.
• Long holds. The longer it stays, the more it reads as a layer.
• Reflective surfaces. The picture will not reflect.

How do you fix a result that came back wrong?

Say which of the three is wrong — the entry, the size or the exit. Because Vizard Agent built those three as separate decisions rather than as one animation, naming the one that bothers you avoids having it redo the other two as well.

• "It slides, it does not get placed." The entry motion changes.
• "Too small." A stated fraction of the frame instead.
• "It is still there at six seconds." The exit moves.
• "Leave it up the whole time." A fixed badge instead.

How does Vizard Agent compare to doing it yourself?

By hand an image gets a position keyframe and a slide, because those are the controls in front of you. The result is correct and lifeless, and the usual response is to add an easing curve, which makes a smoother slide rather than a placement.

Common questions

Why does a slide look wrong? Because nothing in a real room moves at a constant speed in a straight line.

Which corner works best? Bottom left or bottom right. Vizard Agent picks one to suit the shot.

How big should it be? Big enough to read. Tell Vizard Agent a fraction of the frame.

Can it add a shadow? Vizard Agent can, and on a strongly lit shot that helps a great deal.

Does it work on a moving camera? Less well. Vizard Agent will say when the shot is fighting it.

Can the photo stay for the whole video? Yes, though Vizard Agent treats that as a badge rather than a placement.

Will it cover the action? Not if you tell Vizard Agent what has to stay visible.

Can it come back later? Yes, though Vizard Agent treats a second entry as a different moment.

What if the photo is the wrong shape? Vizard Agent crops rather than shrinking it.

Does it check the exit? Yes. Vizard Agent checks it on frames, which is where leftovers show.

Can it hold and then pull away? Yes, and Vizard Agent finds that reads better than cutting away.

Will it match the video's grain? Ask and Vizard Agent will match it to the footage.

Can I use several photos? Yes. Vizard Agent gives each one its own entry rather than stacking them.

Why did mine look pasted on? Because it moved like a layer rather than like something a person was holding.