Direct Win32 API, weirdly shaped Windows, and why most of them disappeared

In the past, Windows desktop development meant working directly with the Win32 API, which let you freely change the shape of windows themselves.

Using a message loop and HRGN, you could create non-rectangular windows, and SetWindowRgn let you implement non-standard shapes like oval windows. By removing the title bar and handling drag operations directly, you could design a window's appearance and interaction together.

There was also a method of using a bitmap as the window's region. By making the background color transparent, you could overlap the image with the window boundary, and going further, using WS_EX_LAYERED and UpdateLayeredWindow enabled smooth alpha channel-based transparency and even animation. In other words, the pixels themselves became the shape of the window.

But this freedom came at a high cost.

• Basic features like dragging, resizing, closing, keyboard input, and DPI handling all had to be built manually.
• With the conveniences provided by the window frame gone, maintenance difficulty increased.
• Users also gradually came to prefer simplicity and stability over experimental appearances.

As a result, many apps today have converged on similar forms built on top of web-based wrappers like React, Electron, and Tauri. The article views this trend as a trade-off that sacrifices performance and memory efficiency in exchange for development convenience and ease of deployment.

In conclusion, Win32 remains the environment offering the lowest-level control over desktop windows, and the possibility of recreating weirdly shaped windows still exists. However, this stands as an example that simultaneously shows both that it's technically possible, and how expensive it is to actually turn it into a real product.