This project aims to create an AI-powered model that generates summaries of YouTube videos. The model utilizes both speech-to-text technology and image captioning to extract key information from video, providing users with quick and informative summaries.
- Be aware that the GUI can be slow on MacOS. Some of the fields and buttons may take some time to load.
If you are using a Mac to run this program, you should therefore give the GUI some time to load. We have also noticed that the GUI is not working on certain Mac-computers. We have not been able to find the reason behind this problem, but we think it might be due to some unknown issues with SSL-certificates on some Macs. This problem could be bypassed by adding this to the
S_to_T.py-file, but it is not recommended:
import ssl ssl._create_default_https_context = ssl._create_unverified_context
- When entering a YouTube video into the program, the video will be downloaded on your computer. This includes the video, the audio and the video chunks. Depending on the length of the video you input, you will get an amount of chunks downloaded on your computer. These can be deleted when you have received your summary.
- We recommend that you use BLIP as there is not a great difference between BLIP and BLIP-2 when it comes to the summaries, but BLIP-2 is more computationally heavy. To use the BLIP-2 model, we recommend to have Cuda installed.
- The YouTube video must be in English as the Speech Recognizer only supports English.
- Download the code from the zip-file or clone the GitHub repository:
git clone https://github.com/sofiejoha/gen-ai.git- It is recommended that you use a virtual environment to install the dependencies. Create a virtual environment:
python3 -m venv venvActivate the virtual environment:
source venv/bin/activate- Install dependencies:
pip3 install -r requirements.txt If your Mac gives you a tkinter-error, try installing the package with brew:
brew install python-tkYou may also specify your Python version, for example:
brew install [email protected]- Run the project:
python3 gui.py- Enter the url of the YouTube video you want summarized. Please note that the video should be in English.
- Choose between BLIP and BLIP-2. Note that we recommend that you have Cuda installed for BLIP-2.
- Choose the personality of the summary. The summary is written differently depending on which personality you want the agent to have. Do you want it to sound like your university professor? Or maybe you want a mafioso to summarize the video for you? Are you a beginner and want it explained as if you are a child? No problem, just choose the kindergarten-personality!
- Click the "Summarize"-button and wait as the program processes your video.
If you want a short video to test the program with, this is a video explaining the C programming language in 100 seconds: https://www.youtube.com/watch?v=U3aXWizDbQ4
