SummarAI employs the combination of PyMovie and Whisper to provide video transcription services, allowing users to ask questions about the video's content with the help of Llama2 and LlamaIndex for more engaging and insightful interactions.
SummarAI is divided into three different files:
-
1_extract_audio.py: This file utilizes the PyMovie library to extract audio from the specified video in a matter of seconds. -
2_transcribe_audio.py: In this case, the "speech recognition" model from OpenAI, known as Whisper, is used to transcribe the audio into a text file. -
3_llm_over_data: Finally, this file makes use of LlamaIndex, which, based on the Llama2-13B model, provides context to the data from the previously prepared video and responds accurately within that context.
To run the SummarAI, you can use pipenv. First, ensure you have pipenv installed. Then, navigate to the root directory of the project, where you can install the dependencies and activate the virtual environment using the following commands:
pipenv install
pipenv shellAfterwards, simply run the file you want using one of these commands:
python 1_extract_audio.py
python 2_transcribe_audio.py
python 3_llm_over_data.py✅ Extract audio from videos
✅ Transcribe audio using Whisper
✅ Resolve your inquiries with the assistance of Llama2
If you encounter issues while installing the llama-cpp-python library, you may need to run the following commands:
set CMAKE_ARGS=-DLLAMA_CUBLAS=on
set FORCE_CMAKE=1If the error persists, please check your C compiler (gcc) or, if you are using Windows, update/install the "Desktop development with C++" feature by utilizing the Microsoft C++ Build Tools