OpenScreen 2

OpenScreen is a free, open-source screen recorder and video editor for Windows, macOS and Linux that takes a raw screen capture and turns it into a finished demo. Its stated purpose is that a raw capture goes in and a finished demo comes out, in the category Screen Studio defined. It is aimed at anyone who records their screen — a founder demonstrating a product, a developer walking through a tool, a support person showing a workflow — and wants the recording to look finished without learning a separate editing suite. Both halves of the job live in one application: the recorder captures the screen, and the editor places zooms, restyles the cursor, dresses the frame, and cuts the video, including cuts described in writing and handled by an AI agent over MCP. Screen recordings usually look unfinished on the first take. The cursor is small, the frame is a bare rectangle of desktop, and the moment that matters — a button being clicked — scrolls past at normal speed. The usual fix is to re-shoot, or to move the file into an editing program and rebuild the framing by hand. OpenScreen's approach is to make the recording look good before you touch anything: a finished frame with wallpaper, padding, shadow, rounded corners and a little motion blur is applied by default, and zooms are planned from where you clicked. It is MIT licensed, with no watermark and no account, and it continues the original OpenScreen project, which its creator archived after v1.5.0. Recording is done with the operating system, not around it. You pick a window or a display, and OpenScreen uses the capture path the system itself provides: macOS goes through ScreenCaptureKit, Windows through Windows Graphics Capture, and Linux through PipeWire and the ScreenCast portal. System audio is captured without an extra driver, and Windows uses Media Foundation for webcam capture. The important technical detail is that the pointer is recorded as data rather than burned into the pixels, which is the only reason the cursor can be restyled later instead of being committed at record time. On macOS, cursor shape and clicks become available once Accessibility is granted. The editor is built around a small number of things you will actually do. For style, you can swap the background to an image, a colour or a gradient behind your recording with no re-shoot, set the background animation to none, Drift, Aurora or Waves, and apply background blur. For effects, padding, motion blur, shadow and roundness each composite live, so you frame it your way. For the cursor, size, smoothing, motion blur and click bounce are all adjustable, because every move should read on screen. Five cursor packs ship with the app — Studio Ink, Prism Glow, Pop Coral, Pixel Candy and Star Sprout — and each can be drawn as a 3D object with a shadow. Zoom placement is the headline editing behaviour. Zooms from your clicks means each take is zoomed where you clicked, by a fixed rule, with no AI and no network call, and it can be turned off in Record mode. On top of that sits a 3D camera: each zoom can stay flat, turn the screen left or right, or orbit it while following the zoom's focus, and depth of field softens the far side to add depth. Around the screen you can place a frame — a window, laptop, phone or screen, in light or dark — over an animated background. On the timeline, zooms, speed ramps, trims and comments each land as a pill, so one click places every zoom and the edit stays visible rather than hidden in menus. The transcript is a second editing surface. You edit video like text: delete a word or a silence and the cut lands on the timeline, and nothing is destructive, so the original material stays intact. Captions lean on the same idea. Transcription runs on your machine because whisper.cpp ships with the app; the model downloads once on first use, and after that it works with the network off across 100 languages. The audio never leaves the laptop, and what comes back is editable text — set the typeface, the size, the colour and the position, then burn it into the render. Webcam work is handled with layouts and segmentation. There are four layouts: picture in picture, dual frame, vertical stack, or no webcam at all. Full Camera gives a stretch of the timeline to the camera, so it grows to full screen and then eases back. Blur, cut out or replace is driven by a segmentation model that separates you from the room on your own computer, and nothing is uploaded. The same take can be framed for any aspect ratio — 16:9, 9:16, 1:1, 4:3, 4:5, 16:10 or 10:16. Blur regions, smooth or mosaic, as a rectangle, an oval or a freehand shape, can be drawn over an email, a key or a password before the clip is shared. Then there is the agent. You describe the edit and the agent makes it. The wizard places zooms by watching where your cursor went; the agent goes further by reading the actual transcript and the actual timeline, so it answers with timecodes you can go and check, telling you which spans it will cut and how much that saves. Every edit it makes is an ordinary undoable one, it needs a provider key you supply, and it is off by default — nothing runs until you connect one. Export writes the file on your machine, which is where the local-first promise is easiest to see. Output is MP4 from 720p up to source resolution, at 24, 30 or 60 frames per second, in H.264, or a GIF. The encode runs on your machine and counts frames while it does, so you can see progress rather than wait behind a queue. There is no account, no watermark, and the file is on disk when the bar fills. The outcome for users is a demo that looks deliberate without a second tool. Nothing is uploaded: recording, transcription and rendering all happen on your machine, and your video never leaves it. Text leaves only when you ask — the chat panel and caption translation each use a key you supply, and transcription downloads its 264 MB Whisper model once, on first run. Every feature ships free for personal and commercial use, with no paywalls, no premium tier and no usage caps. Because transcript cuts and agent edits are non-destructive, you can experiment without risking the take. The concrete workflows are the ones the product was built around. You record a product demo and let the zooms fall where you clicked, as in the walkthrough showing a soil sensor being paired in the Fern garden app: no app store, no account, the probe talks to the laptop, and the whole thing took forty seconds — with the dead air still to be trimmed. You trim that dead air by deleting words and silences in the transcript instead of scrubbing a timeline. You cut a vertical or square version of the same take for a social post without recording again. You mask a key, an email or a password with a blur region before sharing a support clip. You let the agent read the transcript and propose which spans to cut, with timecodes you can verify. And you burn subtitles in whatever language the audience needs. OpenScreen runs on Windows, macOS and Linux from one source tree with native capture on each. Distribution is broad for a free tool: a Microsoft Store listing, a .dmg, a .exe, a .deb, a .rpm, a .pacman, an AppImage and a Nix flake, with a signed and notarized macOS build — the Windows .exe is not code-signed, so SmartScreen warns before it runs. The interface is available in 15 languages: English, Arabic, Czech, German, Spanish, French, Italian, Japanese, Korean, Russian, Turkish, Vietnamese, Brazilian Portuguese, Simplified Chinese and Traditional Chinese. The tech stack leans on the platform: ScreenCaptureKit, Windows Graphics Capture, PipeWire, Media Foundation and whisper.cpp for local transcription, with H.264 encoding on the way out. The takeaway is straightforward. OpenScreen is a free, open-source screen recorder and video editor that makes the first take look finished, keeps the whole pipeline on your machine, and leaves the tedious part of editing — placing zooms, trimming dead air, cutting spans — to rules, transcripts and, if you want it, an AI agent you connect with your own key. No watermark, no account, MIT licensed, and every edit reversible.