Image Generation from text using Artificial Intelligence .

-image generator

An image generator is a software that creates images from the prompt text that can be used many purposes such in Graphic designing , Book templates etc. Many companies will use Image generator for creating new designs for anything they want. Many people will also use image generators for more traditional purposes, such as creating memes and creating artworks.

-why to use image generator

Image generators are very useful because they can create an unlimited number of images without the need to find models or real
world images. Even sometimes images are purely fictions but it does look like it fictions.

-Types of AI Image Generator
At the time we are writing the post we have only 3 AI Image Generator available which is OpenAI DALLE 2 , Googles IMAGEGEN , Open Source DALL-E -MINI

DALL-E 2 : OpenAI DALL-E 2 is an AI model developed by OpenAI that has been trained to generate images from text . Its has more than 12-billion parameter which is approax size of 30 GB . It is successor of DALL-E which has only 3 billion paramters. It is more fast in generating Images from DALL-E and also generates more clear images than DALL-E .

Model Architecture is given below

Examples:

1. Example

2.Example

3.Example

4.Example

Imagen :
Google Imagen is an AI model developed by Deepmind of google that has been trained to generate images from text comparable to DALL-E-2 outputs , even in some cases it is much better than DALL-E . Its parameter size and model size is not known yet as this model access is given only to google employees . Model Architecture is given below

Examples:
1.Example

2.Example

3.Example

4.Example

DALL-E-MINI :
DALL-E-MINI is AI model developed by open sources . It is developed by the research paper of DALL-E-2 . It has two versions one is mini and one is mega . Mini contains around 300 Million parameter where mega is still training , It has around 1 -2 billion parameters. It has same architecture as DALL-E-2 .

Examples:

1.Example

2.Example

Main Difference between DALL-E-2 and Imagen is model structure and quality produce by each model . As our finding shows as that DALL-E-2 try to follow your order in its own way. But Imagen trys to complete your order as you like .

4. PARTI - Pathways Autoregressive Text-to-Image

PARTI is also AI model developed by Google inc . It uses Autoregressive method for developing rather than transformers which was used by Imagen . It gives really exciting and innovative images than Imagen . Its highest variant model have 20 Billion parameter . It is not yet released for public use .

Model Architecture :

Example 1:

Example 2:

-Conclusion

There are many different types of image generators. Each of these models has its own pros and cons, making it difficult to choose
which one to use. The best way to choose an image generator is to first understand what you are trying to accomplish.