CANVAS : CREATIVE AUDIO NARRATIVE AND VISUAL ASSEMBLY SYSTEM

Abstract

Artificial intelligence (AI) is reshaping content creation, particularly in video production, by automating complex processes. The CANVAS system harnesses AI to streamline video creation from scripting to visual and audio generation, reducing technical barriers and enabling professional-grade outputs for users with varying levels of expertise. Central to CANVAS capabilities are advanced AI models for text, image, and audio generation. A structured methodology was used to evaluate open-source AI models based on performance, adaptability, and resource efficiency. The models were assessed for their ability to generate coherent text, high-quality visuals, and natural-sounding audio. This evaluation ensured compatibility with CANVAS objectives and guided model integration decisions. The analysis showed that Latent Diffusion Models produce detailed images efficiently, Retrieval-Augmented Generation (RAG) models enhance text relevance, and the XTTS model enables realistic multilingual speech synthesis. These findings help balance output quality with computational requirements, supporting CANVAS goal of providing accessible, AI-powered video production. The system demonstrates the potential of open-source AI models to simplify multimedia content creation while maintaining quality, scalability, and efficiency. Keywords: Artificial intelligence, video production, content automation, CANVAS system, text-to-image, text generation, text-to-audio, Latent Diffusion Models, Retrieval-Augmented Generation (RAG), XTTS, multimedia content, open-source AI, model evaluation, resource efficiency, scalability, professional-grade video.