screenIt is a small, single-binary screenshot and screen recorder for Windows, and an open-source Snagit alternative. Press Print Screen, drag over anything, mark it up and share it. It records to MP4 with microphone, system audio and webcam, and every action can be scripted from the command line or by an AI agent.
No Electron, no webview, no ffmpeg: one native executable that uses the capture, video and OCR components already built into Windows.
| Capture | Region, window, full screen, scrolling, menu, object, multi-area and freehand capture. Crosshair overlay with an 8x magnifier and size badge. Print Screen works even when another tool has claimed the key. |
| Annotate | Arrows, shapes, text, callouts, numbered steps, highlighter, blur, pen, stamps, magnifier, crop and cut-out. Everything stays editable until you export. Unlimited undo/redo. |
| Effects | Borders, shadows, torn edges, reflection, spotlight, watermark, perspective and page curl. |
| Grab text | Pick an area or window on screen, or use Grab text on the image you have open. The text goes to the clipboard, is saved in the library as a text entry (searchable), and a notification confirms it. Offline Windows OCR in English and Arabic, or optionally a local vision model through LM Studio. |
| Smart tools | Smart Redact (emails, phone numbers, card numbers, IPs, URLs, your own terms), background removal. |
| Record | MP4 (H.264) of an area, a window or a whole monitor, up to 4K (native size or scaled to 1080p / 1440p / 2160p; GPU desktop duplication keeps it smooth), with microphone, system audio, webcam picture-in-picture and cursor effects. A floating pause/stop toolbar that never shows up in the video. |
| AI tutorial (alpha) | Describe a task. A local language model watches the live app and takes one step at a time, asks you when it is stuck, then screenIt records the screen and puts the video in the library. Same LM Studio model as Grab text. Still being tested. |
| Edit video | Trim, cut, combine, export to GIF, make a video from images, export MP4. |
| Organise | Library grouped by day with thumbnails, tags and search across screenshots, videos and grabbed text. Saves go to Pictures\screenIt (OneDrive-aware). |
| Languages | 28 interface languages, including full right-to-left Arabic, switchable in Settings. Translations are plain files you can edit. |
| Export | PNG, JPEG, GIF, BMP, TIFF, WebP, PDF, clipboard, or the editable .scit project format. |
| Automate | Every button is a command, reachable from screenit-cli, a local JSON control channel and an MCP server. |
Settings are saved automatically and apply as you change them: the interface language, what happens after a capture (open the editor or save silently), where captures are saved, whether to start with Windows, recording quality, and text recognition.
Under Text recognition choose the engine (Windows OCR, a language model, or Windows first and
the model as a fallback), the recognition language, and the optional LM Studio server: its URL,
model, API key and timeout, with a Paste button beside the URL and key and a Test
connection button. The API key is stored in plain text in settings.json.
screenIt ships with 28 interface languages: English, Arabic, Bulgarian, Croatian, Czech, Danish, Dutch, Estonian, Finnish, French, German, Greek, Hungarian, Irish, Italian, Japanese, Latvian, Lithuanian, Maltese, Polish, Portuguese, Romanian, Russian, Chinese (Simplified), Slovak, Slovenian, Spanish and Swedish. Pick one in Settings > General > Language; it applies immediately.
Arabic is fully right to left: the window layout mirrors, letters join, and English words and
numbers inside Arabic text keep their order. Arabic and English text recognition is built in.
Most non-English translations are machine-drafted and welcome review. They are plain
<code>.toml files: Settings > General > Edit translations opens the folder, and Reload
translations applies your changes.
The AI tutorial tile on the capture bar is an alpha feature. Opening it shows an amber banner: "Alpha release. Still being tested."
Write what the tutorial should teach. screenIt asks the same local model as Grab text (set the LM Studio URL, model and API key under Settings → Text recognition) to look at the live window and choose the next move. A step is kept only when the screen changes. If the model is stuck, it stops and asks: answer with a hint, say you did that part yourself, skip the chapter, or finish there.
Plan writes the script as JSON (tutorial-<time>.json in the output folder, or the path in the script box) and shows a progress bar while it works. Rehearse checks the script on the real app. Record films each chapter from the screen. Produce builds the finished movie, with chapter cards and captions, and reads the recording straight through. Each chapter clip and the finished movie are added to the library as video, tagged tutorial.
The output folder defaults to Videos\screenIt tutorials in your user profile. Leave App to launch empty when the app is already open. With no app path and no window title, Plan uses the frontmost window you left open.
A step-by-step guide is on the wiki: AI tutorial.
Record with Auto-polish on, then open the recording in the editor. Tick what you want
(zoom on clicks, cut idle, speed up typing, shortcut badges), pick the frame shape and press
Render polished video. It saves clip-polished.mp4 next to the recording, and needs nothing
extra installed.
Silence cuts and noise reduction also work out of the box. Subtitles, filler-word cuts and narration voices need a local speech server (whisper.cpp or Speaches); setup and every setting are in docs/audio.md.
Grab screenit-v0.2.34-windows-x64.zip from the Releases
page. Either run the .msi (per-user install, no admin rights, starts minimized with Windows; pass STARTUP=0 to skip that) or unzip the portable zip anywhere and run screenit.exe. Turn Start with Windows on or off in Settings; screenit.exe --minimized starts straight into the tray. The MSI is unsigned, so SmartScreen may warn.
Requirements: Windows 10 (1903 or later) or Windows 11, 64-bit.
| Action | How |
|---|---|
| Capture | Print Screen, or the big Capture button. Drag a region or click a window. Esc or right-click cancels. |
| Full screen | Ctrl+Shift+F |
| Grab text | Grab text tile or top-bar button, then drag an area; or Grab text in the editor to read the open image. The text is copied and saved in the library. |
| Record | Record tile. Toggle Microphone, System audio, Webcam and Cursor effects first, then drag an area, click a window, or press Enter for the whole monitor. |
| AI tutorial | AI tutorial tile (marked Alpha). Write what to teach, then Plan, Rehearse, Record, and Produce. Stop with Ctrl+Shift+Pause or by moving the mouse. |
| Editor tools | V select, A arrow, L line, R rectangle, E ellipse, T text, S step, H highlight, B blur, P pen, C crop |
| Undo / redo | Ctrl+Z / Ctrl+Y |
| Export PNG | Ctrl+S, or Export PNG in the top bar |
| Zoom | Ctrl+wheel, Ctrl+0 fit, Ctrl+1 actual size |
| Back to the capture bar | Ctrl+Shift+B |
Closing the window keeps screenIt running in the system tray, so Print Screen keeps working. Use Quit in the tray menu to exit. Launching it again just brings the running window forward.
You need the Rust stable MSVC toolchain and the Visual Studio Build Tools with the Windows SDK.
git clone https://github.com/Busbar2297/screenit
cd screenit
cargo build --release -p screenit -p screenit-cli
.\target\release\screenit.exeCreate the portable release folder and zip (release/screenit-v<version>-windows-x64/):
cargo xtask package # portable zip into release/
cargo xtask msi # also builds the MSI (downloads the free WiX Toolset v3 on first use)screenit-cli commands --ids # every command id
screenit-cli exec capture.fullscreen # run one command
screenit-cli run '[["capture.fullscreen",{}],["doc.export",{"path":"shot.png"}]]'
screenit-cli mcp # MCP server over stdioThe running app also listens on a token-protected loopback channel. See docs/control-protocol.md.
screenIt is a layered Cargo workspace. The engine is headless and fully tested; the UI is a thin
egui shell on top of it, and all Windows-specific (and all
unsafe) code lives in one crate, screenit-platform.
apps/ screenit (desktop app), screenit-cli, screenit-fixture (test window)
crates/ geom raster codecs doc ops algo vector compose format capture record
library platform engine ui-egui automation testkit
xtask/ layers, ci, parity, scorecard, panic-hunt, perf, package
More detail: architecture, development, audio and speech, known gaps, feature checklist.
Issues and pull requests are welcome. Before opening a PR:
cargo fmt --all
cargo clippy --workspace --all-targets -- -D warnings
cargo test --workspace
cargo xtask ci # add --e2e for capture/recording changes (needs a desktop session)AGENTS.md describes the project rules (no panics in non-test code, unsafe only in
the platform crate, every feature is a command with tests).
Licensed under either of Apache License, Version 2.0 or MIT license at your option. Third-party notices: ATTRIBUTION.md.
Snagit is a trademark of TechSmith Corporation. screenIt is an independent project and is not affiliated with or endorsed by TechSmith.





