Welcome to CyberReader, your personal cyber assistant designed to boost your productivity! Unleash the power of Large Language Models (LLMs) to read and comprehend virtually anything! 📖📚🎧

Done: ✅ Todo: ❓

Conda (Recommended):

conda create -n cyberreader python=3.10
conda activate cyberreader
pip install -r requirements.txt

Pip:

pip install -r requirements.txt

python main.py

Enter the URL or File Path, which is the path of the media or document you want to read. Supported inputs are listed above.

mv config.yaml.template config.yaml

TODO: Implement multi-token storage

Supported Tokens

Fill the Token in the config.yaml (recommended) or enter the token in the GUI.

The API provider will be identified based on the model name.

Change the whisper model used. Generally, the larger the model, the more accurate the recognition, but it will also be more time-consuming.

Supports using GPU to accelerate whisper model inference.

If this parameter is ignored, the language will be automatically detected. For higher recognition accuracy, users can specify the source language in advance.

The desired output language of the user, this parameter will be applied to the prompt to change the final output language.

Media file: tmp

Original text: result/${timestamp}/original.txt

Result text: result/${timestamp}/result.txt

English Video -> English Text -> Japanese Summary

Japanese Video -> Japanese Text -> Chinese Summary

MIT License

Many of my courses require watching a large number of videos for pre-class preparation and completing assignments based on the video content. However, watching videos is often inefficient, and the contents are not easily retained. Therefore, we hope that this project can enhance the efficiency of watching videos and memory retention of the content, significantly improving the efficiency of information retrieval when doing assignments.

Yes, we are planning to support text reading. However, due to academic obligations, we haven't had the time to implement this feature. If you are interested, we welcome you to submit a PR.

Yes, we are planning to support dialogue as well, but it hasn't been implemented yet for the same reasons above.