User friendly

[pszemraj]

user-friendly

• 💖 overall major rewrites to codebase to improve user experience / make it easier to understand things
• ✨ add arguments for calling the main transcription script
• ♿ : remove extra scripts, vid2cleantxt/transcribe.py now main script for transcription
• ⚡ greatly streamlined the process of converting video -> audio
• ✨ add support for audio files in addition to video files
• 🐛 fixed issue where the large model was not using the attention mask
• 🗃️ ♿ updated examples, and 🔥 large video files from repo itself (can still be downloaded with added scripts)
• 💡 📝 added tons of docstrings to explain what functions do (thanks, copilot!)
• ✏️ 🎨 cleaned up and modified code to be more readable, and more!

[pszemraj]

Issue to fix #1 - if user does not have NVIDIA drivers installed, checking for cuda causes crash

see colab notebook here

[pszemraj]

issue to fix #2 - need to synchronize which tensors are passed to CUDA and which are not

see colab notebook for this issue here

[pszemraj]

Issue to fix #1 - if user does not have NVIDIA drivers installed, checking for cuda causes crash

see colab notebook here

resolved, see same link

[pszemraj]

issue to fix #2 - need to synchronize which tensors are passed to CUDA and which are not

see colab notebook for this issue here

resolved, see same colab link for test results