A Visual Novel Generator that transforms plain text into an engaging and visually interactive novel, powered by Open Source Generative AI.
Visual novels are a form of digital interactive fiction that combine storytelling with visuals and audio elements. This project leverages the capabilities of open-source generative AI to turn mundane text into a visually captivating and auditory experience for users.
We've incorporated two major models into this project:
llama-2-13b-chat.ggmlv3.q2_K.bin
stable-diffusion-xl-base-1.0
Thanks to 4-bit quantized models, we can efficiently run the latest Language Model models like Llama 2 13B on GPUs with less than 16 GB VRAM.
Follow these steps to get started with the Pygame VN Generator:
-
Clone the GitHub repository or download it:
git clone https://github.com/abhiram1809/VN_Game_Generator -
Install the required libraries:
pip install -r requirements.txt -
Host and run the following two notebooks on a GPU instance. Ensure that both notebooks collectively use less than 30 GB of VRAM. Make sure you have an NGROK account set up for Auth tokens.
- Once the servers are up and running, add the file you want to convert into a visual novel to the main directory. By default, it will run 'The Bet.txt', you can use other formats like .docx and .pdf as well. Change the following line in VN_pygame.py to your file name and ngrok llama and SD servers:
colab_ngrok_url_llama = "http://de5f-34-125-218-8.ngrok-free.app"
colab_ngrok_SD = "http://87c3-34-29-116-85.ngrok-free.app/"
file_path = "The Bet.txt"
- Run VN_pygame converter.py:
python 'VN_pygame converter.py'