Skip to content

Latest commit

 

History

History
36 lines (27 loc) · 1.51 KB

README.md

File metadata and controls

36 lines (27 loc) · 1.51 KB
Raw

VideoToAriticles

This project is based on the ideas from the blog An attempt to solve the LLM challenge by @karpathy and attempts to implement the VideoToArticles functionality.

Note: Progress is being made on VideoToArticleWithNoImage, with the image insertion feature currently in development.

How To Run

Set your OpenAI Key in the .env file, and be sure choose the model which support long-context, otherwise the it will failed.

Based on your device, setting the env blew:

  • device: cuda or cpu, Apple Sillcon please choose mps.
  • whisper_model:
    • tiny
    • base
    • small
    • medium
    • large
  • LONG_CONTEXT_MODEL:
    • if True, it will choose a prompt that intergrate previous paragraph as a reference to generate the next paragraph. This will increase the contextual coherence of the generated paragraphs.
    • if False, it will choose a shorter prompt which saving token consumption
  • ARTICLE_LANGUAGE:
    • cn or en. Based on your video language to choose prompt language

After you set your config, run the following command:

  1. pip install -r requirement.txt
  2. cd VideoToArticleWithNoImage
  3. python main.py
  4. Enter your local video file path

If the program runs smoothly, the final generated articles will be stored in the result folder. The temp folder is used to store intermediate files during the program's execution, which can be deleted upon successful completion.

TODO

  • Intergrate image in articles