Reelmator — Screen recording for Unprofessionals

00 — The session you know

Take 37.

Developers aren’t directors, and nobody hired a cutter. That’s why every screencast dies the same death: one perfect run, thirty-six almost-perfect ones. Press the button for the classics — and what Reelmator does about each of them.

01 — The cursor track

The pointer is never filmed.

Reelmator captures the cursor as a separate data track and draws it synthetically when the film is rendered.

After a cut, the cursor is interpolated from the previous click to the next — takes don't have to be perfect.
Raw paths during drags, no jump at cuts.
Pointer size, style and motion are render settings — decide them after the recording, not before.

The same cut, filmed vs. drawn.

02 — The stage

A stage, not your desktop.

The recording is a composed stage: the apps you put on it, on the project backdrop, at the project ratio. Everything else on your Mac is never captured — no desktop cleanup, no leaked notifications.

Stage apps — the capture filter includes exactly the apps you stage; the rest of your desktop does not exist on film.
Backdrop — a wallpaper of your choice behind the stage, per project.
Ratio — every clip is composed to the project canvas; one setting for the whole series.
Stage display — optionally the display itself runs at the canvas ratio while recording: menus and popovers lay themselves out inside the frame, the notch is letterboxed away, and everything returns to normal on stop.

03 — Editing & camera

Nondestructive editing.

Recordings are never modified. Cuts, zooms and retiming are render instructions — reversible weeks later.

The zoom camera frames clicks and interactions, not pointer motion.
Automatic targets, adjustable per keyframe.
Optical-flow slow motion.

Re-Record

One edit. Every language.

Reelmator replays the recorded session in your app — AX anchors, resolved through its string tables, time-locked to the original take. Hit Re-Record for French and Reelmator does the rest: switches your app to French, resets its preferences to the captured snapshot, relaunches it, loads the demo document fresh — and performs the whole take again. The same film in French, edit intact.

Step 01

Record the session once

Clicks, menus, scrolling and typing are stored as events alongside the video.

Step 02

Cut it once

Trim, zoom, title, score — one edit for all language versions.

Step 03

Re-Record in any language

Anchors resolve against the localized UI; the app performs the take again; the edit stays.

One edit — all six languages come out of this.

04 — Narration

Narration, per block.

Narration is written and voiced per text block, so you fix one sentence without re-recording the whole track.

Synthetic voices from Google, Apple's on-device engine, or plugins such as ElevenLabs.
Or record your own voice, block by block, with the built-in teleprompter.
Translation runs against the UI glossary of the app on screen — Claude or Apple's on-device models — so the narration says exactly what the interface says. No invented menu names.

Translation uses your string tables.

05 — Finish

Titles, sound, slow motion, export.

Titles

Template-based title cards: font, animation, crossfade.

Click & key effects

Optional: ripples and UI sounds, rendered from the recorded events.

Ambient music

Looped beds, ducked under narration.

Slow motion

Optical-flow retiming, down to 0.2×.

Still insert

Hold any frame for as long as the narration needs — the camera stays live.

Pan & Zoom

Pinch to zoom, drag to pan, keyframed on the timeline; unpinned, the camera follows the cursor.

Export

One project, every language version, rendered from the same edit.

06 — Extensible

A plugin is one file.

Voice and translation services are single JavaScript files with an .echoservice extension. Double-click to install. No bundles, no installers, no restart.

// ElevenLabs.echoservice
service = {
  name: "ElevenLabs",
  capabilities: ["tts"],
  settings: [{ key: "api_key", secure: true }],
  async synthesize({ text, voice }, s) { /* http(…) */ }
};

07 — MCP-first

Built to be driven by an agent.

Every operation is an MCP tool: record_start, timeline_split, clip_rerecord, export_clip and the rest. An agent can produce a finished screencast end to end.

Batch a release's demo videos from one instruction.

Screen recording
for Unprofessionals.