How to work with a phone clip of a laptop screen as your only footage
Send the clip and say what has to be readable. Vizard Agent hunts for the seconds where something legible actually happens on the screen, tests crop after crop to isolate that region and any logo, and lays a subtle dark scrim under the text so it holds up on a phone with the sound off.
What is the short version?
A phone pointed at a laptop is the worst source there is — glare, angle, a moving hand, a screen inside a screen. It is also frequently the only thing that exists, and one clip of a real lesson beats a stock shot of a stranger.
- Go to Vizard Agent with the clip.
- Say what has to be readable and who it is for.
- Say it will be watched muted, scrolling.
What do you need before you start?
The clip and a clear idea of the one thing a viewer must see. That last part does the most work: a poor source cannot show everything well, so choosing what it has to show well is the decision the whole edit hangs on.
- The clip. As filmed.
- The must-see moment. The thing that proves it.
- The destination. Where the viewer ends up.
- The frame. 4:5, 9:16, or both.
- Sound on or off. Assume off unless you know.
What do you type into Vizard Agent?
Describe the source honestly and name the constraint. "Handheld phone filming a laptop screen" is not an apology — it tells Vizard Agent to solve for legibility first rather than reaching for a look the footage cannot carry.
Prompt
Variants worth knowing:
- "Sound off." Changes every decision that follows.
- "This has to be readable." Names what the crop protects.
- "Both frames." One primary, one secondary.
What does Vizard Agent actually do?
Here is the order Vizard Agent worked in on a real ad cut from a single handheld clip. Steps four to six are the search that makes the rest possible: it is looking for evidence, not for pretty shots.
- Inspects the source's properties and extracts frames across it.
- Builds contact sheets and reviews the footage as a whole.
- Extracts detailed frames from the sections that look usable.
- Finds the timestamps where a red circle is actively being drawn.
- Locates every moment with drawing marks on the shared page.
- Extracts frames around that action and reviews them closely.
- Extracts the logo tile frames and checks the dimensions.
- Tests crop candidates and views the options.
- Tests framing for the page and the logo tile separately.
- Generates tighter logo crops, then tighter still.
- Centres on the logo tile and checks it at the delivery ratio.
- Measures the source's loudness.
- Installs the typeface the captions need.
- Generates the subtitles and a subtle dark scrim for legibility.
- Test-renders the scale and subtitles and reviews a montage.
- Refines the scrim and the caption styling.
- Renders the primary 4:5 export and QCs it as a contact sheet.
- Renders the 9:16 version and QCs that separately.
Step four is the insight worth stealing. In a clip of somebody teaching over a video call, the moment that proves the product is the moment the teacher marks the page — so the edit is built around finding those seconds rather than around the best-looking framing.
Steps eight to eleven look repetitive and are the job. A screen filmed at an angle has one small region that is actually readable, and finding it is a matter of trying crops and looking, not calculating one.
What does the result look like?
An ad where the readable part of the screen fills the frame, the marking action is the beat the cut is built on, and the captions sit on a scrim dark enough to survive the glare behind them.
It still looks like a phone filming a laptop, and that is fine — for a service sold on being real people teaching real lessons, the roughness is evidence rather than a flaw.
When does this not work well?
There is a hard floor set by the source itself. Where nothing on the screen resolves at any crop Vizard Agent tries, no amount of clever framing produces a readable shot, and the honest answer is to capture a screen recording instead of rescuing this one.
- Nothing legible anywhere. Re-capture as a screen recording.
- Heavy glare across the screen. The reflection moves with the hand.
- Very low resolution. Cropping in has nowhere to go.
- Fast camera movement. Every frame is blurred.
- Personal data on screen. Say so; it needs covering, not cropping.
How do you fix a result that came back wrong?
Name the thing you still cannot read on screen. Vizard Agent keeps every crop test it ran, the moments it located in the clip and the scrim settings, so tightening a frame or darkening a scrim becomes a re-render rather than a fresh search through the footage.
- "I still can't read it." Cropped tighter from the tested candidates.
- "The captions vanish." Scrim darkened under them.
- "Wrong moment." Re-cut to a different marking action.
- "The logo is cut off." Re-centred at the delivery ratio.
How does Vizard Agent compare to doing it yourself?
By hand this is doable and dispiriting: scrub for the readable seconds, try crops, discover the text is still soft, try again. The temptation is to give up and cut to stock footage, which loses the only thing the clip had going for it.
| By hand | Vizard Agent | |
|---|---|---|
| Finding the usable seconds | Scrubbing | Searched for by what happens on screen |
| The crop | One or two attempts | Many, compared on frames |
| Legibility | Hoped for | Scrim added and tested |
| A second ratio | Re-do the crops | Rendered from the same analysis |
| The fallback | Stock footage | Your real clip, framed properly |
Common questions
Is a phone clip of a screen usable at all? Often, yes. Vizard Agent finds the part that is and builds around it.
Would a screen recording be better? Usually, and Vizard Agent will say so. But a live lesson carries something a recording does not.
What is a scrim? A subtle dark layer under text so it stays readable. Vizard Agent tests how dark it needs to be.
Can it remove the glare? Partly. Vizard Agent will tell you when the reflection moves too much.
Can it straighten the angle? Somewhat, at the cost of resolution. Tell Vizard Agent which matters more.
Will it add stock footage? Not unless you ask. The point here is usually the real clip.
Can it deliver two ratios? Yes. Vizard Agent renders both from one set of crop decisions.
How does it find the right moment? Vizard Agent looks for what happens on screen — a mark drawn, a page turned.
What about the audio? Measured and kept, but assume it plays muted and caption everything.
Can it blur personal details? Yes. Tell Vizard Agent what is on screen that should not be.
Does it upscale the crop? It can, though a soft source stays soft. Vizard Agent says when.
Can I supply a better clip later? Yes. Vizard Agent carries the same decisions across to it.
Is it worth reshooting? Sometimes. Vizard Agent will tell you what it could not rescue.
Why not just use stock? Because a stock shot of a stranger at a laptop proves nothing, and a real lesson with a marked page proves the thing you are selling.