Study Transcriber

This is a work in progress. The goal is to automatically transcribe lecture videos using Whisper, improve the transcripts using some local LLM, and embed the results into a vector database. This can then be used to easily access lecture contents for review.

Install packages:

# Pytorch, e.g.:
pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu118
pip install langchain pypdf fpdf sentence_transformers chromadb cryptography==3.1

Run semantic search

python ./semantic_search.py [OPTIONS] ./path/to/input_folder/

Options:

• --types [type1] [type2] [...]: File types to index. The only supported types are pdf, txt. Default is pdf txt.
• --lang [language]: Language for stopword removal in the query. This is only relevant for the highlighting. Use the full language name, like english, german, etc. Supported languages: arabic, azerbaijani, basque, bengali, catalan, chinese, danish, dutch, english, finnish, french, german, greek, hebrew, hinglish, hungarian, indonesian, italian, kazakh, nepali, norwegian, portuguese, romanian, russian, slovene, spanish, swedish, tajik, turkish.
• --print: Also print the results to stdout.

Note that highlighting is always done in green color (as yellow is too mainstream) and is still experimental.