Harnessing Stable Diffusion Model for High-Resolution Text-to-Image Synthesis

Harnessing Stable Diffusion Model for High-Resolution Text-to-Image Synthesis

Dr. Vegi Fernando A

Assistant Professor dept. CSE-AIML Dayananda Sagar

University

Ashish Patil

ENG21AM0013

dept. CSE-AIML

Dayananda Sagar University

Atharva T

ENG21AM0014

dept. CSE-AIML Dayananda Sagar University
Diksha Sinha

ENG21AM0033

dept. CSE-AIML

Dayananda Sagar

University
Tenzin Ludup

ENG22AM3011

dept. CSE-AIML Dayananda

Sagar University

Abstract—This study focuses on turning written descriptions into high-quality pictures using powerful AI diffusion models. These models use iterative denoising, which begins with a noisy image and gradually refines it to produce realistic and coherent outputs that match the user-provided text.Pre-trained models, such as Stable Diffusion, are used for their efficiency in text- toimage generation. Fine-tuning on specialized datasets improves adaptability, allowing the system to handle a wide range of textual inputs, from straightforward descriptions to complicated prompts. Techniques such as latent space processing maximize computing efficiency while maintaining output quality.With an impressive 90 percent accuracy A U-Net architecture that incorporates attention processes enhances the model’s capacity to generate detailed and accurate pictures.

Index Terms—AI Diffusion Models,Text-to- ImageGeneration,Stable-Diffusion,Latent Space Processing,U- Net

Architecture,Attention                                   Mechanisms,Image Synthesis,Generative

AI,Frechet Inception Distance (FID),Visual Content Creation´