Neverclick brings computer vision keyboard control to Windows

Lazo Velko has released Neverclick, a free Windows utility that lets users trigger mouse-style actions from the keyboard by placing shortcut labels across the screen. The app is notable less for another keyboard productivity pitch than for Velko’s claim that it avoids Windows accessibility APIs and instead uses a small computer vision system to read screen pixels.

Neverclick is available now for Windows at no charge. Velko says it requires no account, runs offline, serves no ads and does not collect user data. The project has a GitHub repository, but the readme says it is currently used for issue reporting rather than as a full source-code release.

The utility is designed to let users click buttons, menu items, application icons, window controls and specific points inside text without reaching for a mouse. According to Velko’s documentation and comments, it can also select multiple targets for repeated clicks, such as closing several windows. Drag and drop and text highlighting are not yet supported.

Computer vision instead of accessibility APIs

Velko says he built Neverclick after a repetitive strain injury made mouse use difficult. On the Neverclick website, he says he recovered years ago and still uses the software daily.

The technical claim behind the app is that it does not depend on the UI Automation accessibility system in Windows. In a Hacker News discussion, Velko wrote that he had found accessibility APIs “clunky, slow, and unpredictable,” and said computer vision avoids those issues. In an email to The Register, he said he built the computer vision system himself and did not rely on third-party libraries.

Velko also claims the approach is faster than Windows UI Automation on older machines. He wrote on Hacker News that some users run Neverclick on 10-year-old hardware and report that its computer vision runs immediately while UI Automation feels slow. Those performance claims have not been independently benchmarked in the information available.

According to Velko, the app takes about 40 MB of disk space. He says memory use is about 200 MB on a typical 1080p display, with the number changing by resolution and other factors. He also says Neverclick does not run in a constant loop, so CPU use should be limited when the tool is not being called.

How it works in practice

The Register’s US editor Avram Piltch tested the app on a Windows PC. By default, Ctrl + Enter activated a left-click mode, while Alt + Enter activated Ctrl + left click. Right click is not mapped by default, though users can configure it.

Once activated, Neverclick covers the display with two-letter labels that it calls hints. Typing the matching label moves the cursor and clicks that location, whether the target is a button, a menu entry or a word. Word hints click the middle of a word by default, and users can type a number to target a specific letter. The app also includes shortcuts for switching windows and can show hints either in the active window or across the screen.

The test also found practical friction. The overlay can obscure menus and icons, making it hard to see the underlying interface. Dismissing the overlay requires a click. The default Ctrl + Enter shortcut can also collide with existing application shortcuts, including Gmail’s send command.

Neverclick is Windows-only for now. Velko told The Register that a macOS version is planned, but he has not set a timeline because he still has Windows features to finish, including window dragging and text highlighting through two selected hints.

This story draws on original reporting from The Register.