DEV Community

dark
dark

Posted on Originally published at darkmeteor.hashnode.dev

Turning raw iPhone and Mac screen recordings into 1080p app demos offline

Recording a 15-second screen clip of a new iPhone or Mac feature is fast. Turning that raw .mov into something clean enough for a changelog or social post usually takes much longer: trimming the Control Center pull-down at the end, cutting out a four-second loading pause, and zooming into the small button you tapped so phone viewers can actually read it.

I built TapZoom to do that cleanup on a single screen, running natively on iPhone, iPad, and Mac without uploading footage to a cloud editor.

TapZoom editor on macOS showing a framed screen recording with a smooth zoom keyframe, tap ripple, and timeline cues

How do zoom keyframes work when iOS hides touch coordinates?

On iOS, app sandboxing prevents one app from reading raw touch coordinates recorded inside another app. Instead of pretending to capture hidden system touches, TapZoom runs a fast on-device frame-difference pass over the video (ignoring the top status bar clock) and places visual Screen Change Cues along the timeline wherever the UI updates. You can jump straight from cue to cue, step frame by frame, and tap directly on the preview canvas to drop a zoom highlight (Snappy, Smooth, or Gentle) and a high-contrast tap ripple. You can also add an optional top headline or per-step caption for muted feeds.

What happens to zoom timing when you cut or speed up a middle pause?

In a basic timeline, deleting a dead pause in the middle of a clip breaks the timing of every zoom keyframe you placed after it. TapZoom models edits as piecewise-linear playback segments: whether you trim the head and tail, cut a middle section out, or speed a waiting interval up to 2x or 4x with pitch-preserved audio, every later zoom keyframe, tap ripple, and caption shifts automatically so it stays locked to the exact frames you anchored it on, with multi-step Undo and Redo.

What formats and limits should you expect?

The app is a ~3.2 MB native Swift 6 binary built directly on AVFoundation and CoreGraphics with zero third-party SDKs, running offline on iOS/iPadOS 17+ and macOS 14+ with zero video uploads or telemetry. It handles variable frame rate ReplayKit recordings and converts Display P3 and HLG clips cleanly to Rec.709, exporting deterministic 30fps or 60fps 1080p H.264 MP4 files or transparent HEVC with Alpha (.mov) overlays across 9:16, 4:5, 1:1, 16:9, or Full borderless aspect ratios (with an optional adaptive 9:41 strip to cover the red recording pill on portrait iPhone clips).

Two practical boundaries worth knowing up front: framed and zoomed exports are designed for social posts, launch pages, and changelogs rather than raw unframed App Store Connect App Previews. And while exports on the built-in 8-second sample recording are unlimited and your first rendered 1080p export of your own recording is free with zero watermarks, unlimited new edits and additional recordings use a single one-time Universal Purchase across iPhone, iPad, and Mac, with no subscriptions or accounts.

Overview & App Store download (iPhone, iPad & Mac)

Top comments (5)

Collapse
 
userxnom profile image
Abdullah Naseer •

The piecewise-linear playback model is the detail that sold this for me — cutting a four-second loading pause in the middle of a clip without breaking every zoom keyframe after it is exactly the kind of thing that is easy to get wrong in a basic timeline, and shifting the keyframes, tap ripples and captions together so they stay locked to their anchored frames makes sense. I also liked the honest workaround for iOS hiding touch coordinates: a frame-difference pass that places Screen Change Cues along the timeline, ignoring the status bar clock, instead of pretending to capture hidden touches. A 3.2 MB Swift 6 binary with zero third-party SDKs and 1080p H.264 or transparent HEVC with Alpha exports is a tight footprint for that. One question: when a Screen Change Cue fires for something small, like a loading spinner or a badge count changing, how do you keep the cue list from getting noisy — is there a threshold for how big a frame difference has to be?

Collapse
 
dark_meteor profile image
dark •

Good question. There are three safeguards keeping small UI noise from flooding the timeline:

  1. Block-level threshold + adaptive baseline: instead of firing on raw pixel diffs, the scanner scores regional blocks against both a minimum floor and the clip's own median activity level. Small localized changes like a badge count flipping or a compact loading spinner stay under the floor and produce zero cues.
  2. Peak selection + spacing suppression: adjacent frame spikes get collapsed to their local peak with non-maximum suppression and a per-clip cap, so full screen or sheet transitions always take priority over smaller movements nearby.
  3. Honest boundary: a large half-screen skeleton shimmer can still clear the threshold and leave a few spaced markers, which is also why cues stay strictly non-destructive jump targets on the timeline rather than auto-inserting zooms into your export.
Collapse
 
robertpaschal profile image
Odinaka Robert Nnamani •

Covering the red recording pill with an adaptive strip is a small touch that saves a lot of re-recording. Offline and no uploads makes it an easy tool to trust with raw clips.

Collapse
 
dark_meteor profile image
dark •

Thanks! That red pill ruined way too many otherwise good takes for me too. It samples the clean header background per frame (and backs off on busy photo headers), so if a dark sheet slides up mid-clip the 9:41 bar shifts color with it instead of staying stuck on one flat shade.

Collapse
 
robertpaschal profile image
Odinaka Robert Nnamani •

Per-frame sampling is a neat fix. Does it hold up on busy photo headers, or does the back-off show?