International Research Journal of Engineering and Technology (IRJET)
e-ISSN: 2395-0056
Volume: 12 Issue: 03 | Mar 2025
p-ISSN: 2395-0072
www.irjet.net
AI-Based Text-to-Image Generative Application Sachin Meshram1, Rushikesh Suryawanshi2, Bhavika Salunkhe3, Atul Wasnik4 , Sakshi Waware5 1 Professor, Dept. of Information Technology, Kavikulguru Institute of Technology and Science, Ramtek
2-4 Student, Dept. of Information Technology, Kavikulguru Institute of Technology and Science, Ramtek
---------------------------------------------------------------------***--------------------------------------------------------------------1.2 Objective of the Study
Abstract - Artificial Intelligence (AI) has made significant
advancements in generative models, particularly in the field of text-to-image synthesis. DALL·E, developed by OpenAI, is a state-of-the-art model that can generate realistic and creative images from textual descriptions. This paper explores the working principles of AI-based text-to-image models, their applications in various domains such as marketing, design, and medical imaging, and the challenges they present, including ethical concerns, biases, and computational costs. We also discuss the future scope of AI generative models, highlighting potential improvements in realism, control, and ethical AI frameworks. This study provides insights into how AI is transforming digital creativity and the potential risks and benefits associated with these technologies.
This paper aims to:
2. LITERATURE REVIEW
Key Words: AI Text-to-Image, Generative Models, DALL·E, Deep Learning, Computer Vision, Ethical AI, Artificial Intelligence, Image Synthesis.
The field of AI-driven image generation has evolved significantly over the past decade. Early approaches in computer vision relied on Convolutional Neural Networks (CNNs) for image recognition and synthesis. However, the introduction of Generative Adversarial Networks (GANs) by Ian Goodfellow in 2014 marked a breakthrough in generating realistic images from random noise.
1. INTRODUCTION The rapid development of Artificial Intelligence (AI) has enabled machines to generate images from textual descriptions, opening new possibilities for creative industries, healthcare, education, and digital content creation. Text-to-image generative models leverage deep learning techniques to interpret human-written text and convert it into realistic or imaginative visuals. DALL·E, a neural network developed by OpenAI, has demonstrated remarkable capabilities in generating high-quality images from textual prompts.
Later, advancements in Transformer-based architectures and self-supervised learning led to models that could generate images from textual descriptions. Key developments include:
1.1 Importance of AI Text-to-Image Technology AI-driven text-to-image generation is revolutionizing several industries:
Marketing & Advertising: AI-generated visuals are used in digital marketing, ad campaigns, and social media content creation. Design & Art: Designers use AI to generate unique artwork, concept designs, and product prototypes. Medical Imaging: AI assists in creating medical visualizations, helping in diagnostics and training.
Gaming & Entertainment: AI-generated images contribute to game asset creation and virtual reality experiences.
© 2025, IRJET
|
Impact Factor value: 8.315
Explain the underlying working principles of AIbased text-to-image models. Highlight the key applications of DALL·E in various industries. Discuss the limitations, challenges, and ethical concerns associated with AI-generated images. Provide insights into the future scope and advancements in text-to-image generation.
|
2015 – Deep Convolutional GANs (DCGANs): Improved stability in GAN training for generating images. 2018 – BigGAN: Introduced large-scale image generation with enhanced realism. 2020 – CLIP (Contrastive Language–Image Pretraining): Developed by OpenAI, CLIP enabled AI models to understand textual descriptions and match them with images. 2021 – DALL·E: OpenAI introduced DALL·E, a transformer-based model trained to generate images from textual prompts using GPT-3 and CLIP techniques. 2022 – DALL·E 2: Improved resolution, text-image coherence, and photorealism in AI-generated content. 2023 – Stable Diffusion & MidJourney: Opensource and commercial models that enhanced accessibility and creativity in AI-generated art.
ISO 9001:2008 Certified Journal
|
Page 60