ANIMED-AI: AGENTIC RAG-DRIVEN REAL-TIME ANIMATION INTELLIGENCE SYSTEM

Abstract

Existing e-learning platforms generally provide the same static content to all students without considering their knowledge level, emotional state, or learning pace. This often leads to lower engagement and uneven learning outcomes. This paper presents ANIMEd-AI - an adaptive learning system based on Retrieval-Augmented Generation (RAG) that generates personalized educational animations in real time for each student query. When a student enters a concept or code, the system processes the input, retrieves relevant information from a ChromaDB vector store using LlamaIndex, and uses an AI model to generate a Manim animation script. This script is then rendered into a new MP4 video that visually explains the concept or shows step-by-step code execution. A DeepFace-based webcam module analyzes the student’s facial expressions at regular intervals and maps the detected emotion to changes in content difficulty. An Arduino Uno with RGB LEDs provides simple real-time visual feedback based on the detected emotional state. The system also includes accessibility features: visually impaired students are supported through OpenAI TTS-based audio narration, hearing-impaired students through MediaPipe-based sign language display, and specially-abled learners through a simplified adaptive interface. Evaluation shows that RAG-based grounding improves factual accuracy, emotion-aware adaptation supports better learning performance, and animation-based explanations are easier to understand compared to text responses. ANIMEd-AI combines animation-based learning, code visualization, and adaptive interaction within a single educational platform.