ladaapp/lada

Lada is a tool for restoring pixelated or mosaic regions in videos. It is specifically designed for adult video content where censoring is applied as a blocky mosaic overlay, and it uses AI to fill in and reconstruct those blocked areas frame by frame. The result is not perfect, as the README notes that quality varies by scene and some results may look worse than the original, but the overall aim is improved visual clarity. The tool can be used through a graphical interface or a command-line interface. With the graphical version, you open a video file and can either watch the restoration play in real time or export a new video file to watch later. The command-line version takes an input file path and processes it, with additional options available through a help flag. Both interfaces are available on Linux and Windows. Running Lada effectively requires a dedicated graphics card. The README recommends at least 4 to 6 gigabytes of GPU memory for most content, and real-time playback requires a fairly powerful machine. Without a capable GPU the software will run, but very slowly. Higher-resolution content such as 4K video also demands significantly more system RAM than 1080p content, because frames are buffered in memory during processing. On Linux, the recommended installation method is through Flathub, which provides a packaged version with no manual setup. A Docker image is also available for command-line use, though it requires an NVIDIA GPU and the NVIDIA Container Toolkit. Windows users can download a standalone archive from the project's releases page on Codeberg and extract it to get the application executables directly. The project is hosted on Codeberg rather than GitHub for the release files, though the source code repository is available. GPU support covers NVIDIA cards from the RTX 20 series onward and Intel Arc cards, depending on the installation method used.