Auto zoom screen recording explained
How click-based auto zoom works: separate cursor capture, click events, spring easing, dwell times, motion blur, and when to override it.
Auto zoom is the effect that made apps like Screen Studio popular: you record your screen at full size, and the finished video glides in towards whatever you clicked, holds there while you work, then eases back out. It looks like someone edited the video by hand with a virtual camera, but the recorder does most of it for you. This article explains how that works under the hood, why some auto zooms feel smooth and others feel seasick, and when you should take over and adjust them yourself.
The basic idea
A normal screen recording is a flat video of your whole display. Auto zoom treats that video as a large canvas and puts a virtual camera in front of it. The camera can move (pan) and get closer (zoom). The output video is what that camera sees.
To decide where the camera should go, the recorder needs to know two things that a plain video doesn't contain: where the cursor was and when you clicked. So recorders that do auto zoom capture that information as data alongside the video.
Step one: capture the cursor separately
On macOS, apps capture the screen with Apple's ScreenCaptureKit framework, available since macOS 12.3. Its stream configuration has a showsCursor setting that controls whether the cursor appears in the captured frames. It's on by default.
Recorders built for polished output turn it off. Screen Studio's FAQ describes the approach plainly: it "records your screen, ignoring the mouse cursor, and then adds the cursor back in the final video". While recording, the app logs the cursor's position many times per second, which cursor shape was showing (arrow, text beam, pointing hand), and every mouse-down and mouse-up event.
Keeping the cursor out of the pixels has several benefits:
- Smoothing. The recorded path can be filtered so small hand tremors and jerky movements become a clean curve.
- Restyling. The cursor can be drawn bigger, in a different style, or hidden when idle, all after the fact.
- Click effects. A ripple or circle can be drawn at the exact click position and time.
- Sharp zooms. When the camera zooms in 2x, a cursor baked into the video would be scaled up and blurry. A redrawn cursor stays crisp at any zoom.
Step two: turn clicks into zoom targets
With a list of clicks and their positions, the recorder builds a zoom plan. The naive version would be "zoom on every click", but that produces a camera that never sits still. Better implementations do a few extra things:
- Clustering. Clicks that happen close together in time and space are grouped into one zoom. Filling in three fields of a form becomes a single zoom on the form, not three separate jumps.
- Lead time. The camera starts moving slightly before the click, so the viewer arrives at the button as you press it, rather than after.
- Dwell time. After the last click in a cluster, the camera holds for a moment before zooming out. Without dwell, the camera snaps away while the result of the click is still loading.
- Follow. While zoomed in, the camera can pan gently to keep the cursor in frame as you move within the area, instead of staying locked to the first click.
- Zoom level. A typical default is around 2x. Higher levels make text readable but lose context. Some tools pick a level based on the size of the area being worked on.
Typing is a special case. Some tools, such as FocuSee, say their auto zoom "follows clicks and typing sections", so a long text entry also gets a zoom even without clicks.
Step three: move the camera with easing or springs
The difference between a zoom that feels filmed and one that feels like a glitch is almost entirely in how the camera moves between positions.
Linear and eased motion
A linear move changes position at a constant speed, starting and stopping abruptly. It looks mechanical. An ease-in-out curve starts slowly, speeds up and slows down again, which is much better. Its weakness shows when a move is interrupted: if a new click arrives while the camera is still travelling, a curve-based animation has to start over from zero velocity, and you see a stutter.
Springs
Many recorders now use spring physics instead. Apple's WWDC23 session Animate with springs gives a clear explanation of why. A spring is modelled on an object attached to a real spring, so it "feels more natural and believable to our eyes". When a spring is retargeted mid-move, it "uses the velocity it had when it was retargeted as the initial velocity towards its new destination". For a zoom camera, that means a new click redirects the camera smoothly instead of making it stop and restart.
Apple describes springs with two parameters, duration and bounce, and suggests a bounce of 0 when you're unsure, warning that values above about 0.4 "may feel too exaggerated" for interface elements. The same advice holds for zoom cameras. A camera that overshoots and wobbles back is fun once and tiring by the third zoom. Most good presets use little or no bounce and a duration somewhere under a second.
Smoothing the cursor and the camera separately
The cursor and the camera usually get different settings. The cursor should stay responsive, so it uses light smoothing that still follows your hand closely. The camera should be calmer, so it uses a slower, heavier spring. If both use the same fast setting, the whole frame jitters. If both are slow, the cursor lags behind your clicks and looks fake.
Step four: motion blur
In film, fast movement blurs because the shutter is open for a fraction of a second. Screen recordings have no shutter, so a fast pan or cursor flick is a series of perfectly sharp frames that can look choppy, especially at 30 fps.
Recorders add synthetic motion blur to fix this. They render several in-between positions of the camera or cursor for each output frame and blend them, or smear the image along the direction of motion. Used lightly, it makes movement read as continuous. Used heavily, it makes text unreadable during moves. A slider with a modest default is the usual approach. CleanShot X 5.0's Studio Mode, for example, lets you apply motion blur to cursor movement after recording.
When to override auto zoom
Auto zoom is a good first draft, not a finished edit. Here's when to step in:
- Clicks that don't matter. Clicking an empty area to dismiss a menu or focus a window produces a zoom nobody needs. Delete it.
- Important things without clicks. A chart that updates, a notification that appears, a result loading on the other side of the screen. Add a manual zoom there.
- Zooms that are too tight. If the viewer can't see what changed after the click, zoom out a level or extend the zoom area.
- Zooms that are too short. If the camera leaves before the result appears, extend the dwell.
- Too many zooms in a row. Merge them into one longer zoom, or remove every other one and let the viewer breathe.
- Vertical or square exports. A zoom that works in 16:9 may cut off important parts in 9:16. Check each format separately.
The best workflow in most tools is: record, let auto zoom generate the plan, then scrub through the timeline once and fix the handful of zooms that are wrong. That's far faster than placing every zoom by hand.
Tools that do auto zoom
As of October 2026, these tools offer click-based or cursor-following automatic zooms:
| Tool | Platforms | How it describes the feature |
|---|---|---|
| Screen Studio | Mac | Automatic zooms from clicks, plus manual zooms on the timeline |
| CleanShot X | Mac | Smart zooms that follow the cursor (Studio Mode, version 5.0) |
| FocuSee | Mac, Windows | Auto zoom following clicks and typing |
| Screen Charm | Mac | Zooms added automatically after each click, adjustable |
| Cap | Mac, Windows, Linux | Zooms generated from recorded clicks, or drawn manually (Studio Mode) |
| OpenScreen | Mac, Windows, Linux | Auto or manual zooms that follow the cursor (repository archived) |
| Dolly | Mac | Auto zoom from clicks, editable zoom blocks, manual zooms |
They differ mainly in how the motion feels and how much control you get afterwards, so it's worth trying the same short recording in two or three of them.
Dolly, our app, follows the approach described here: it records the cursor separately, redraws it with smoothing and optional motion blur, and turns clicks into zoom blocks on the timeline that you can move, resize or delete, with manual zooms where clicks don't tell the story. It's free to download for Apple silicon Macs.
Summary
- Auto zoom puts a virtual camera over your recording and points it at what you clicked.
- It depends on capturing the cursor and click events as data, separately from the video.
- Good implementations cluster nearby clicks, lead slightly, dwell after the action and follow the cursor while zoomed.
- Springs with little or no bounce give the most natural camera motion, especially when moves overlap.
- Motion blur smooths fast moves but should stay subtle.
- Treat the automatic result as a draft: delete pointless zooms, add zooms where something happens without a click.




