kolijpjpd/CourseTranslator — reverse-engineered prompt

Build me a simple Python app that takes a video or audio file, transcribes the speech, and outputs clean SRT subtitles with very accurate timing. I want it to also translate the subtitles into Arabic, with proper right to left text rendering so the final file looks correct and readable.

Please make it feel fast and practical for real use, with smart subtitle chunking so lines do not appear too quickly on screen. If possible, use a fast Whisper based transcription approach and GLM 4 for the Arabic translation step, since that seems to be the idea here. I also want it to work on both Linux and Windows, with an easy setup flow that installs what is needed, including FFmpeg if possible or at least checks for it clearly.

Keep the run experience simple, ideally just install requirements and run the main script. If anything in the repo is incomplete or unclear, look up current docs online and fill in the gaps in a sensible way.