We have built this project as part of MongoDB GenAI hacakthon in New York. We are actively working on front-end. Planning to release the application soon.

A small Introduction to the project: Our creation - ComicGen, is a RAG-based scene generation model that can create comic-book-style renderings of fan theories and plot extensions. Using MongoDB Atlas Vector Search we vectorize and store a large repository of open-access books, which can be queried by a user for alternative plots or creative story generation. Using Stability AI's Stable Diffusion 1.0, we generate comic strips in a user-specified illustration style. Nomic AI's embedding model on top of LangChain, combined with MongoDB Atlas Vector Search forms the base of our RAG model. This conjuncted with Mistral AI's Quantized Mistral-7B-Instruct-v0.2 taken from Hugging Face generates fine-tuned scripts and scenes. These scenes are fed into the Stable Diffusion model to iteratively generate comic illustrations and rendered using Pillow.

Some Comics generated by our model:

1.Pokemon Manga-style illustration

2. Pokemon-Marvel-style illustration.