r(ed)izz
Team consisting of a Georgia Tech MS systems engineer (Luxtron; PyTorch/C++, RAG/BERT), a Calsoft Go SDE (GCP/Kubernetes/Pulumi), and an LJMU AI student (PyTorch/TF, FastAPI/React).
Project Description
Title: AI-Powered Meme Video Generation System Using Redis
Description:
This project showcases an AI-driven pipeline designed to generate creative and contextually accurate memes using image and text inputs. The system integrates Redis, LLMs (Large Language Models), and a Deep Learning Model to automate the end-to-end process of meme creation and video generation.
———————————————————————-
The workflow begins with an image and prompt that are sent into the system. The metadata related to these inputs is stored in Redis for fast access and retrieval. Redis acts as both a cache and a vector store, storing meme vectors for semantic search and reuse.
The LLM processes the prompt and image to generate a refined or rewritten version of the prompt. This optimized prompt is then passed to a Deep Learning Model, which produces AI-generated video or image-based memes via the AI Generated Video UI.
Simultaneously, Redis stores and retrieves meme vectors and descriptions, allowing the LLM to analyze and enhance meme context. The result is a meme description and corresponding AI-generated output that blend humor, context, and visual creativity.
Key Components:
Redis: High-speed data storage for metadata and meme vectors.
LLM: Handles prompt rewriting, meme understanding, and contextual generation.
Deep Learning Model: Generates video memes based on refined prompts and the User-input Image
AI Generated Video UI: User-facing interface to view and interact with generated content.
Metadata Store: Maintains structured data for input management and retrieval.
Use Case:
Entertainment and Fun!