what is video-pause-remover fr?

Video Pause Remover cuts the dead, still moments out of a video, like frozen screen recordings, silent pauses, or long stretches where nothing on screen moves, and produces a new file that is your original video with those moments removed. Unlike tools that cut based on audio silence, this one watches for visual stillness, checking whether the picture itself changes. That makes it especially useful for screen recordings and timelapses, where someone might talk over a frozen screen and an audio-based tool would keep that dead time in. The analysis runs on the GPU, using MLX and Metal on Apple computers or CuPy and CUDA on NVIDIA graphics cards, with a slower CPU fallback available everywhere else. It works by comparing frames in small tiles to detect real motion while ignoring things like brightness flicker. Once a video is analyzed, the result is cached, so adjusting settings afterward is instant and does not require reanalyzing the file. The tool comes with both a web interface and a command line tool. In the web UI, you can drag a slider to control how strict the motion detection is, watch a visual timeline of where cuts will happen, draw boxes to ignore or focus on specific parts of the frame, and preview the trimmed result before exporting. The command line version supports similar options and can output a JSON report describing exactly what was cut. A key technical detail is how the export works: about ninety percent of the output video is copied directly from the original file without any quality loss, and only small pieces right at the cut points are re-encoded. This means the final video keeps close to the original quality rather than being fully recompressed, which most competing tools do. The README compares it directly to a similar tool called auto-editor and to basic ffmpeg utilities, explaining that this project runs faster on the GPU and preserves more of the original video quality. Setup requires ffmpeg and Python 3.11 or newer, plus a virtual environment for installing dependencies. The project notes that audio during removed pauses is also cut, and that the NVIDIA GPU code path has not yet been tested on real NVIDIA hardware.