Screen recording that zooms in by itself, on Windows

You have seen the videos: a product demo where the screen glides in towards whatever is being clicked, holds, and pulls back out, with a soft shadow around the window and a cursor that moves like it means it. Nobody edited that by hand. It is one feature - the camera follows the pointer - and it is the difference between a demo that looks made and one that looks recorded.

The app most of those videos come from, Screen Studio, is macOS only and has been since it launched. So the question for everyone else is what does this on Windows.

what auto-zoom actually is

A screen recorder captures the whole screen. Auto-zoom is a second pass over the recording: a virtual camera moves across that captured image, scaling in around the places where something happened - usually clicks - and easing back out when nothing does. The recording itself never changes. That is why it can be re-cut afterwards, and why a good tool lets you move, lengthen or delete a zoom after the fact.

It works because it does what an editor would do manually and what a viewer wants anyway: at 1440p, a menu item is about forty pixels tall, and nobody watching on a phone can see which one you picked.

three ways it goes wrong, and what to look for

Auto-zoom is easy to implement badly, and the failures are specific enough that you can test for them in five minutes with any tool you are considering.

1. The drawn cursor lands in the wrong place

Most tools draw their own pointer - bigger, smoother, with a click ripple. To put it in the right place they have to map the pointer’s position, which the operating system reports in whole-desktop coordinates, onto the rectangle that was actually recorded. On a single monitor those are the same thing. On a second monitor, or when recording one window, they are not.

Get it wrong and the pointer is drawn a long way from where it was. In our own tool, before we fixed it, a click that belonged at x = 1632 was drawn at x = 933 on a two-monitor desk. Nothing about the video looks broken - the pointer is just subtly, maddeningly beside the button.

Test: record a window - not the full screen - on a multi-monitor setup, click something near an edge, and watch where the drawn cursor is.

2. The zoom cuts off the thing you clicked

A naive implementation treats the click position as the centre of the zoomed view and clamps it so the view stays inside the frame. Click something in the bottom-right corner - an account menu, a Save button - and the clamp pushes the view back until the very thing you clicked is outside it. The menu that opens is cut off at the edge.

The fix is to treat the click as a point of interest rather than a centre: keep it in shot, and let the view sit flush against the frame edge when there is no room to centre it.

Test: click something in a corner and open a menu from it.

3. The pointer walks out of the shot

This is the one that separates the tools. If zooming is treated as a filter - “move the view some fraction of the way towards the cursor each frame” - then a fast movement outruns it and the cursor leaves the visible area entirely. The viewer is watching an empty region of a toolbar while you do something off-screen.

Treating it as a camera fixes it: a dead zone the pointer can roam in without the view moving at all (ours is about a quarter of the visible width), a damped spring that carries the view when the pointer pushes past it, reading the cursor slightly ahead of where it is, and a hard limit - a leash - that the pointer can never be outside. A fast sweep should also widen the shot rather than chase it, the way a camera operator pulls back when they lose their subject.

We measured this on two real recordings. With the follow-the-cursor approach the pointer was outside the visible area for 5.8% and 13.4% of the zoomed frames. With the camera, both were zero.

Test: record yourself dragging something quickly across the screen, then watch it zoomed in.

the options on Windows

OBS is the right answer if what you need is a recording and you are happy cutting it yourself - it is free, it is excellent, and nothing here competes with it on capture. The reason people pay for something else is the hour per video that keyframing zooms costs, every time.

the rest of what makes a demo look made

Auto-zoom is the headline, but a demo video that reads as finished usually has four more things, and it is worth knowing they exist so you can look for them:

• A backdrop and rounded window frame, so the recording is a window on a surface rather than a raw screenshot of your desktop.
• Click and keystroke sounds timed to what actually happened, which is only possible if the recorder captured input timing while recording.
• Silence cut automatically - the dead air while you think is most of the length of a first take.
• Captions, because a large share of the people who watch a demo do it with the sound off.

Who wrote this. We build owntools, and screeni is the recorder in it - so the three failures above are not hypothetical, they are bugs we shipped and then fixed, with the measurements from fixing them. If you are comparing tools, run those three tests on ours too. Our dictation tool is the free part of the same app.