---
name: "screenlife_editing"
description: "Edit screen recordings into watchable tutorials and demos. Use when asked to polish a screen capture, add zoom cuts, cursor emphasis, callouts, keystroke overlays, or pace a tutorial for retention. Covers zoom levels and easing, cursor smoothing and click highlights, callout graphics, dead-air removal, speed ramps, aspect-ratio reframing, and executable ffmpeg recipes."
---

# Screenlife Editing

## Purpose
Turn raw screen recordings into tutorials people finish: guide the eye with zoom cuts and callouts, make the cursor readable, keep the pace tight, and reframe for any platform. Screen content is unforgiving: static pixels, tiny text, wandering cursors. The edit has to do the watching work for the viewer.

## Workflow
1. **Inspect the recording.** Note resolution, frame rate, dead air, cursor visibility, and the key moments (clicks, reveals, results). Plan edits against the brief: what is the viewer supposed to notice, in what order?
2. **Fix the frame first.** Crop chrome, scale up for delivery headroom, reframe to the target aspect ratio. Details in `references/techniques.md`.
3. **Cut for pace.** Remove silence and dead air, jump-cut on movement or natural speech pauses, speed up typing and repetitive navigation. The first 30 seconds are the steepest drop-off: start with the result or the problem.
4. **Direct attention.** Zoom cuts (150% focus, 200% callout, 300%+ tiny text) with eased transitions, cursor smoothing plus click highlights, callout boxes and arrows on key UI, keystroke captions where shortcuts matter.
5. **Pattern interrupts.** Every 60-90 seconds, change something: zoom cut, graphic, sound effect, lower third. No shot stays visually unchanged longer than a few seconds without a reason.
6. **Verify.** Watch the render at 1x. Check: cursor readable on a phone-sized preview, text legible at delivery resolution, cuts feel invisible, audio has no clipped words.

## Output Contract
- Delivers the finished video file (or an edit plan with exact filter commands when the user edits elsewhere).
- Calls out every intentional choice: where zooms land and why, what was cut, what speed ramps apply where.
- If the source recording is too low-res for the delivery target, says so plainly instead of upscaling past usefulness (recording below ~2x delivery resolution leaves no zoom headroom).

## Operating Rules
1. Never snap-zoom: every zoom moves with an ease in and out (0.6-1.0s), or it reads as a glitch.
2. The cursor never wanders: idle cursors get parked or hidden, narration never points away from the cursor.
3. Callouts clarify, never decorate: one callout per idea, minimal text, never covering the thing being explained.
4. Cut the dead air before adding effects: pacing fixes come before polish.
5. Prefer cutting and reframing over effects plugins. ffmpeg filters in `references/ffmpeg-recipes.md` are tested and preferred for headless/batch work.
6. Respect the platform: 16:9 for YouTube, 9:16 for Shorts/TikTok (reframe, never letterbox as an afterthought), 1:1 for feeds.
7. Follow Anthony's standing style law in any generated on-screen text: no em dashes, no AI-speak.