International Journal of Research Publication and Reviews, Vol 4, no 4, pp 4932-4937 April 2023
International Journal of Research Publication and Reviews
Journal homepage: www.ijrpr.com ISSN 2582-7421
Text-To-Image Generation Using AI
Pavithra V
1
, Rosy S
1
, Srinishanthini R B
1
, Prinslin L
2
1
Computer Science and Engineering, Agni College of Technology
2
Assistant Professor, Computer Science, and Engineering Department, Agni College of Technology
DOI: https://doi.org/10.55248/gengpi.234.4.38568
ABSTRACT
This model is suggested to produce the photos that are provided in the text. This has the power to turn abstract images into actual ones. DALL-E is necessary for
the conversion. Making artistic, realism-based graphics from the description will be enjoyable. Kotline and Java were used to create this Android application
project. For this project, we employed a natural language description prompt. Through prompts, it generates visuals. This will help incorporate our many ideas
and thoughts into the diagrammatic presentation. DALLE can be used for business endeavors such as advertising, publishing, selling, etc. It will display images of
our choice, resulting in anthropomorphic imagery and the collaboration of unconnected thoughts. It is viable and produces believable items. We can turn our
inventive thoughts into reality using this Android application project. It is user-friendly software with no visual problems. And we can't discover this fictitious
picture generator on any search engine. This Android application project will generate an image in whatever size you want. And it does not affect image quality.
A picture's quality size will be 256*256, 512*12, and 1024*1024. Based on the quality of our network, we may select a quality size. We must first ensure that the
network infrastructure is in place before we can use this application.
Keywords: Diffusion Models, Energy-based Models, Visual generation.
I. INTRODUCTION
OpenAI's DALL-E is a ground-breaking artificial intelligence program that can generate high-quality photographs from textual descriptions. This
cutting-edge technology is a game changer in the world of picture production, with limitless applications in advertising, design, and even medicine.
The procedure for employing DALL-E is straightforward. DALL- E creates a comparable picture based on a textual description provided by users. For
example, if you wanted an image of a blue cat playing with a ball of yarn, you would describe it to DALL-E, and it would generate a unique image that
matched your parameters.
DALL-E's technology is built on powerful neural networks and machine learning algorithms that allow it to analyze text and create related visuals in
seconds. The program is always learning and developing, so it will continue to create increasingly realistic and detailed photos over time.
Overall, DALL-E is a power that has the potential to change the way we think about image production and design. Because of its ability to generate
high-quality pictures from simple text descriptions, it opens up new avenues for innovation and creative expression in several enterprises.
1.1 OBJECTIVE OF THE STUDY
Examine the accuracy and quality of the pictures produced by DALL-E using a variety of textual descriptions. Compare DALL-E's performance to
those of other cutting- edge image synthesis models. Examine DALL-E's possibilities in a range of fields, including advertising, design, and the arts.
Discuss the moral ramifications of producing fake pictures that are difficult to tell apart from genuine ones.
II. LITERATURE SURVEY
[1]. "DALL-E: Creating Images from Text" by Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, Scott Gray, Chelsea Voss, Alec Radford, Mark
Chen, Ilya Sutskever, and OpenAI. This paper introduces DALL-E and describes its architecture and capabilities.
[2]. "Visualizing and Understanding DALL-E" by Aditya Ramesh, Mikhail Pavlov, Gabriel Goh, and Ilya Sutskever. This paper presents an
analysis of DALL-E's performance and visualizations of its internal representations.
[3]. "Generating Images from Text using Invertible Generative Networks" by Swami Sankaranarayanan and Aravind Srinivasan. This paper
explores the use of invertible generative networks for image synthesis, including a comparison to DALL-E.