How to put a screen recording into the laptop in a photograph
Have Vizard Agent animate the photograph first and measure the screen's corners on the animated frames. A mask built from the original still will not fit, because adding breathing and hand movement to a photo shifts the laptop slightly — and the mask has to follow the screen that is actually on screen.
What is the short version?
You have one photograph of someone at a laptop and a recording of your app. You want a video of them using it. The screen is at an angle, so the recording has to be warped into it, and the photo has to move or the whole thing reads as a still.
- Send Vizard Agent the photo and the recording.
- Let it animate the person before it fits the screen.
- Expect the mask to be proved with a flat colour first.
What do you need before you start?
The photograph and the recording, and a note about what should move. "Blinking, breathing, small natural hand and body movement" is the right level of ambition — anything more and the pose changes, which breaks the composite you are about to build.
A green or plain screen in the photo helps enormously. It gives Vizard Agent an unambiguous region to key, and it means a leftover edge can be measured rather than argued about.
What do you type into Vizard Agent?
Describe both halves in one message, and tell Vizard Agent the screen fit must be exact. People usually ask for the insert and forget to ask for the movement, and a perfectly composited screen sitting inside a frozen photograph looks worse than either of the two would on its own.
Put this video inside the laptop screen in this photo, matched exactly to the screen's size and angle. Add small natural movement to the person — blinking, breathing, slight hand and body motion.
The word "angle" is what rules out the wrong approach. A screen seen from the side is a four-corner problem, and scaling a rectangle to fit it will always look pasted on.
What does Vizard Agent actually do?
Vizard Agent works in the order that survives contact with reality: animate the photo, then measure the screen, then composite into it. The animation has to come first because it changes the geometry, and measuring before it simply means measuring the wrong thing.
In that session the sequence ran:
- Read the screen's area and angle from the photograph, and checked the recording's format and length.
- Built a loop of micro-movement on the photo with the laptop and the composition held as still as possible.
- Warped the recording to the screen's four corners rather than scaling it, testing two coordinate orientations instead of guessing.
- Validated the placement with a plain colour field, to separate a position problem from a video problem.
- Overlaid a red mask on the photo to confirm the green area was covered exactly, then measured the overlay's boundary pixels from that test frame.
- Isolated the alpha channel and viewed it over white to check the polygon's real shape.
Then the discovery that reorders everything: the animation had moved the green screen's edges slightly, so the mask built from the original photo no longer fitted. The four corners were re-measured on the animated frames and the mask rebuilt from those.
What does the result look like?
A short video of a person at a laptop with your app running on it. They blink and shift their weight, the screen content plays, the screen's edges stay where they belong, and nothing about the insert announces itself to a viewer.
Leftover green gets measured rather than judged by eye. A thin line at the edge of a screen can be spilled key colour or it can be the bezel's own reflection, and those need opposite fixes — so the pixels get read at working resolution rather than assessed from a thumbnail.
When does this not work well?
When the animation moves the laptop too much. Generated movement is not controllable to the pixel, and a version where the person leans will drag the screen with them past what a fixed mask can follow. The fix is less movement, not a cleverer mask.
A screen at a very steep angle is also hard. Past a certain obliqueness there are too few pixels across the far edge for the content to be legible, and the honest answer is a different photograph.
And a photo with a reflective or switched-on screen already gives you something to remove before you can add anything. A plain or green screen in the original is worth arranging if the photo is still to be taken.
How do you fix a result that came back wrong?
If green shows at an edge, say which edge. Vizard Agent measures the pixels there to establish whether it is key spill or the bezel, and only then decides whether to grow the mask or leave it alone.
If the screen content looks stretched, the corners are wrong rather than the scaling. Ask Vizard Agent to re-measure them on the animated frames — that is the failure this whole ordering exists to prevent, and it recurs if the animation is regenerated.
If the person's movement looks unnatural, ask for less of it. Almost every problem in this composite gets better with a smaller animation, including the ones that look like masking faults.
How does Vizard Agent compare to doing it yourself?
In a compositing tool this is a corner-pin track, and on a still photograph it is one of the easier jobs there is — until you animate the plate. Then the pin needs to follow, and the movement was generated rather than shot, so there are no tracking markers and nothing to solve against.
Measuring the corners on the animated frames sidesteps that entirely. The other half is the validation habit: putting a flat colour in the hole first, then a coloured mask, then the actual video, so that when something looks wrong you already know which of the three it is.
Common questions
Does the photo need a green screen? It helps a lot. Vizard Agent can work from a plain screen too, and a lit one is hardest.
Why animate before masking? Because animating moves the screen. Vizard Agent measures the corners afterwards for that reason.
How much movement is safe? Blinking, breathing and slight hand motion. More than that drags the laptop with it.
Can it match the screen's angle exactly? Yes, by warping to four corners rather than scaling. Vizard Agent measures them from the frame.
Will the audio from my recording survive? Yes. Vizard Agent measures the finished track to confirm it is intact.
What if there is still green at the edge? Vizard Agent reads the pixels to tell spill from the bezel before changing anything.
Can it loop the movement? Yes. Vizard Agent loops it so the photo stays alive for as long as the recording runs.
How long can the video be? As long as your recording. Vizard Agent loops the photo animation underneath it.
Can I use a phone instead of a laptop? Yes, though Vizard Agent has fewer pixels to fill, so the content must be simpler.
Does the person's face change? It should not. Vizard Agent checks the animation has not altered the photo's structure.
Can I see the mask on its own? Yes. Vizard Agent can output the alpha over white, which is where shape problems show.
What if the screen is partly hidden by a hand? Say so. That needs the hand kept in front, which is a second mask.
Is this cheaper than filming it? Usually, and for Vizard Agent it is the only route when you have one photograph.
Will it look generated? Not if the movement stays small, which is the one thing worth keeping under control.