tarteelai/tarteel-ml ? reverse-engineered prompt

Build me a Python project for working with the Tarteel dataset, focused on downloading the data, preprocessing audio, and preparing it for model training.

I want scripts that can download the dataset, create train, validation, and test splits, generate the Quran letter and vocabulary files, and build a CSV in the format needed for speech recognition training. It should also include the basic training and comparison utilities, plus a simple demo area for trying things out. Make the scripts easy to run from the command line with help text, and assume the user may need to install ffmpeg and ffprobe first.

Keep it organized like a real machine learning repo, with clear preprocessing steps, reusable helper code, and a clean setup so someone can follow the README and get from raw data to training ready files without much friction. If you need to check current library docs online, go ahead and do that.