Focalet Capture

Share selected screen regions with the editor, terminal or chat app you already use. Capture runs independently of Desktop and needs no agent sign-in.

Download and open

Choose Capture for your platform in the download table:

Platform Install Where to find Capture
Windows 10/11 x64 Run the Capture setup executable Start menu and system tray
macOS 12+ Apple Silicon or Intel Open the matching DMG and drag Focalet Capture into Applications Applications and menu bar
Ubuntu 24.04 x64, GNOME Wayland Install the Capture .deb with sudo apt install ./Focalet-Capture-Ubuntu-amd64.deb Applications and GNOME panel

Windows packages include .NET. Mac packages include the browser context helper. Ubuntu's package manager installs the native dependencies. Flutter and agent runtimes are not required. Windows installers are unsigned; Mac apps are ad-hoc signed and DMGs are not notarized.

On Mac, allow Screen Recording to select pixels and Accessibility to include context and paste. The menu's Permissions item opens these settings. On Ubuntu, choose Enable desktop integration on first launch. If GNOME has not loaded the extension, sign out and back in, then reopen Capture. Capture reuses a current Focalet extension; its independent installer does not overwrite files owned by Desktop. Screen sharing is authorized when you invoke capture.

Both apps can be installed together. On Windows and Mac, quit Desktop or change its capture shortcut before opening Capture. On Ubuntu, Alt+A routes to Capture while it is running, and returns to Desktop when Capture quits.

Capture, then choose where to paste

  1. In the source window, press Shift+Alt+A (Shift+Option+A on Mac).
  2. Drag up to eight regions. Ctrl enters continuous selection; choose a drawing tool to annotate. Pen, arrow, rectangle, ellipse, colors and undo are available. Choose Done or press Enter.
  3. Switch to your destination and click its input at the intended caret.
  4. Press Alt+A (Option+A on Mac).

Paste follows image A → context A → image B → context B. Images remain separate at their selected dimensions. Capture never submits the message. Successful pastes finish silently. Press the paste shortcut again to reuse the batch, including in a different app. Cancelling a new capture preserves the previous batch.

Capture does not guess a previous destination. Paste uses the currently focused window, stops on focus or clipboard changes, and never retries interrupted steps automatically. Known protected fields are excluded. Destination accessibility checks inspect identity and state, not input text. The receiving app controls whether it accepts images and where attachments appear.

On Windows, attaching an image can briefly move focus to the page or an upload control. Capture waits up to 1.5 seconds for the original editor to regain focus, using fresh accessibility state after each paste. A rich editor rebuilt in place can continue only when it uniquely replaces the original editor in the same container. Capture never moves the caret or sends an accepted image again. A different input, container, window or replaced clipboard still stops the sequence. Typing, navigating, clicking or scrolling during paste also stops it. Presses while pasting do not queue another batch; briefly held shortcut keys pause the current sequence until released.

The system tray, menu bar or GNOME panel provides Capture, Copy last batch, Copy text, preferences, and Quit. Text-only and slower-image preferences are in the menu on Windows/Mac and Preferences on Ubuntu. Windows uses the same smooth rounded menu design as Desktop, with keyboard navigation and native accessible actions and checkboxes.

Each automatic image step offers only image formats: PNG/DIB/RTF on Windows, PNG/TIFF on Mac, and PNG on Ubuntu. GNOME owns Ubuntu clipboard transfers so the receiving input keeps focus; Capture does not activate a hidden window to write the clipboard. Its following text step contains the region label, readable context and complete bounded snapshot JSON. Inputs that ignore images still receive the text. DOM, accessibility IDs, hierarchy, geometry, state, source and alignment remain available when captured reliably.

Default minimum waits are 0.5 seconds per image and 0.15 seconds per text step, plus a brief period after clipboard reads. Slower image paste allows at least three seconds for apps that read a preview and fetch the image again later. A clipboard read cannot prove that the receiving app finished uploading an attachment. Use Text only to skip images entirely.

Copy last batch offers a rich HTML document with separate images and complete plain-text fallback (also RTF on Windows). Manual paste depends on the receiving app's format support. No combined image is created.

Capture retains batches in memory. Each region is limited to 32 megapixels and the images together to 32 MiB. Quit releases shortcuts, capture helpers and the batch. The app captures only after invocation. Frozen selected pixels are preserved; changed or unaligned source structure is reported as image-only.

Build and verify

From a committed checkout on the target OS:

python3 scripts/package_capture.py linux
# macOS: python3 scripts/package_capture.py macos
# Windows: python scripts/package_capture.py windows

Build prerequisites and regression suites are in CONTRIBUTING.md. Capture's native CI is independent of Flutter and the Rust agent broker. It checks image/context order, Unicode metadata, alignment, package inventories and installers. Windows tests include native editors, Chromium and an isolated Electron receiver. Ubuntu tests run real capture and ordered paste in a private GNOME Wayland desktop. Mac checks compile both architectures, verify native clipboard formats, selection geometry and the mounted DMG; OS permission and actual destination-app acceptance remain separate from these checks.

See product use cases and the component map.