This is the largest release ScreenBuddy has had — 295 commits since 1.6.0. A recording is no longer one clip you decorate; it is a project you direct.
New
Projects and the storyboard. Your work is now a saved project rather than a session you must not close. The timeline has a storyboard lane of scenes you can reorder by dragging, duplicate, delete, and give their own entrance, exit and backdrop. A scene can hold an imported video or image, or be a text card with no footage at all — and you can record a new scene straight into an open project.
Templates. A gallery of starting points, browsable on one axis, each with a live preview on hover. Pick one and it composes real, editable regions into your project — nothing is locked afterwards.
The Lens. A physically-modelled depth of field. Aim a focus point with an on-screen reticle, or let it follow the cursor; set the aperture and the size of the sharp area independently. The backdrop steps back and dims with the shot, and the preview and the export agree frame for frame.
Callouts. Lift a rectangle of your recording out of the frame as a card, with a headline, borders, glow, gradients, backdrop blur and reveal animations. On the 3D device frame it lifts off the screen as a real object and swings with the lid.
Real 3D camera work. Every device frame — none, the flat bezel and the MacBook — now renders through one 3D camera, so orbit, dolly, pan and focal length mean the same thing whichever you pick. A shot can hold as many keyframes as you like, each with its own easing, and a zoom can pull the camera out as well as in.
Captions on Windows and Intel Macs. Local transcription shipped for Apple silicon only, silently. Both are now built, verified against a real sample at build time, and the packaging step fails rather than shipping a platform that promises captions it cannot deliver.
Sound and motion. Click and keyboard sounds baked from the recording itself, selectable sound packs with audition, and animated backdrops.
A new look. The whole interface has been regraded onto a cinematic design language: dim, desaturated chrome so the video is the only saturated thing on screen.
Fixed
Windows: the recording bar's menus sat on a black slab. Every dropdown on the HUD opened inside a large opaque rectangle. The same bug framed the webcam bubble in a black square and squared off the source picker's rounded corners. All three are transparent now.
Windows: the installer is signed. It no longer says "Unknown publisher" — the prompt names the publisher. Windows may still show a SmartScreen warning until the download reputation builds up; that part is time and volume, not something that can be configured or bought.
"Source" export quality was minifying your recording. It named the export canvas, not the recording — so the padding inset scaled a 1920×1080 recording down to 1588×893 before encoding, 68% of its own pixels. Source exports are now genuinely 1:1. The encoder also moved off WebRTC rate control, which was costing quality on Windows for no benefit.
Auto-update reconnects. 1.6.0 shipped without the update manifest despite claiming otherwise, so no installed copy could see a new version. This release carries it for both platforms; installing it once restores automatic updates.
Windows parity. Windows cursor artwork instead of macOS shapes, the editor chrome macOS already had, and window recordings no longer capture the whole desktop.
Autosave actually saves. A guard latched on first run and never released, so the project system persisted nothing.
Plus more than a hundred other fixes across export audio, undo, captions, licensing and the timeline.
For AI / MCP users
apply_edits gained multi-beat camera moves, the lens vocabulary, callouts and per-scene region anchoring. Cuts now merge when they touch, so upsertRegion {kind:'trim'} can return an id other than the one just created — get_guide and the tool description both say so. Templates are listable. As always there is no export tool, and there will not be one: the AI builds the edit, a human reviews it and ships it.