OpenScreen is a free, open-source screen recording and video editing software for Windows, macOS, and Linux, designed to turn raw recordings into polished presentation videos. It is licensed under MIT, with no paid tiers, no usage limits, no watermarks, and no account registration required, and it is free for both personal and commercial use.
Purpose and Positioning The project continues the original OpenScreen project (archived by its original creator after v1.5.0) and is positioned as an all-in-one "record plus edit" tool: input a raw capture, output a fairly complete presentation video. It emphasizes local processing: recording, transcription, and rendering all happen on the user's own machine, and videos never leave the local device. Only when the user actively uses the chat panel or subtitle translation does text leave the device via a key provided by the user.
Main Capabilities
- Background replacement: place images, solid colors, or gradients behind the recorded footage without reshooting.
- Framing: provide padding, motion blur, shadows, rounded corners, and other effects, all composited in real time.
- Cursor editing: adjust cursor size, smoothness, motion blur, and click bounce effects. Because the cursor is recorded separately as data rather than burned into the pixels, its size and style can still be changed after recording.
- One-click zoom: automatically place zooms based on click positions, using fixed rules, without AI and without making network requests; it can be disabled in recording mode.
- Timeline editing: each edit, such as zoom, speed ramp, trim, and comments, lands on the timeline as a "capsule"; it supports deleting words or silent segments like editing text, and cuts land on the timeline and are non-destructive.
- 3D camera: each zoom can stay flat, turn the screen left or right, or orbit while following the zoom focus; depth of field can soften the far end.
- Cursor packs: five cursor packs are provided, Studio Ink, Prism Glow, Pop Coral, Pixel Candy, and Star Sprout, each of which can be drawn as a 3D object with shadows.
- Borders and motion effects: window, laptop, phone, or screen borders, with light or dark options, and backgrounds can be paired with Drift, Aurora, or Waves animations.
- Four layouts: picture-in-picture, dual frame, vertical stack, or no camera.
- Full camera: a segment of the timeline can be handed over to the camera, with the picture zooming to full screen and then easing back.
- Person segmentation: a segmentation model runs locally to separate the person from the room, supporting blur, cutout, or replacement, and content is not uploadedTransmission.
- Any aspect ratio: A single recording can be framed as 16:9, 9:16, 1:1, 4:3, 4:5, 16:10, or 10:16.
- Area blur: Use smooth or mosaic blur on email, key, or password areas; shapes can be rectangles, ellipses, or freehand shapes.
Screen Recording and Export Recording uses the operating system's own capture path: macOS via ScreenCaptureKit, Windows via Windows Graphics Capture, Linux via PipeWire and the ScreenCast portal. The pointer is recorded as data, not burned into pixels. System audio requires no extra drivers. Export to MP4, with resolution from 720p to source resolution, frame rate 24, 30, or 60, encoded as H.264, and GIF output is also possible; encoding is done locally and frames are counted, with no queueing, no account, no watermark, and once the progress bar finishes, the file is already saved to disk.
Subtitles and Transcription whisper.cpp is provided with the application, and the model is downloaded once on first use (about 264 MB), after which it can work offline. Audio does not leave the local machine, and what is returned is editable text: you can set the font, font size, color, and position, then burn it into the rendered result. About 100 languages are supported.
AI Editing Agent The wizard places zooms by observing cursor position; the agent goes further: it reads the actual transcript and the actual timeline, returns verifiable time codes, and explains which clips will be cut and how much time will be saved. Every edit made by the agent is a normal, undoable operation, and the user must provide their own vendor key; it will not run until connected. Off by default.
Platforms and Installation One source tree, native capture on each platform. Microsoft Store listing, .dmg, .exe, .deb, .rpm, .pacman, AppImage, and Nix flake are provided. Windows also has an .exe installer, but it is not code-signed, so SmartScreen will warn first: choose "More info," then choose "Run anyway." The macOS version is signed and notarized, and after granting accessibility permissions it can obtain cursor shape and clicks. Linux requires PipeWire and xdg-desktop-portal. All artifacts are on the Releases page, and complete installation steps are in the Installation documentation.
Interface Languages 15 interface languages are provided: English, Arabic, Czech, German, Spanish, French, Italian, Japanese, Korean, Russian, Turkish, Vietnamese, Brazilian Portuguese, Simplified Chinese, Traditional Chinese.
Limitations and Notes The transcription model requires downloading about 264 MB on first run, after which it can be used offline. The AI editing agent requires the user to provide their own vendor key and is off by default. The Windows .exe installer is unsigned and may trigger a SmartScreen warning. Linux screen recording depends on PipeWire and xdg-desktop-portal. Text leaves the local machine only when the user actively uses the chat panel or subtitle translation, and the user must provide a key. Apart from that, video, audio, and transcription are all processed locally.
