Now there is a growing trend among tech savvy netizens to create AI systems capable of generating images based on a brief description. They are called as text to image synthesis and image to image synthesis to visualize and generate any kind of artwork can anyone think off and enter the commands in prompt. With these systems, you write something that you want to see in a painting, and the IA searches on the Internet for these words and searches associated
images to it. Once it has the images the system creates a new composition joining fragments from the images that it had trained before.
There are many AI diffusion model that you can use to create new pictures. We will further talk about only DALL·E and DreamStudio.
DALL-E 2 is a new AI based image synthesis system (created by OpenAI) that can create realistic images and art from a description in natural language. With this system, you can edit images in a realistic way from a natural language caption. For example, you could add or remove elements taking into account shadows or reflections.
DreamStudio is the front end of StableDiffusion (created by stability ai) and it has a nice interface with a simple user design. Introducing the text query on the text field the AI will generate the image. You can adjust some parameters such as the width or the height of the image too.
Comparison of image synthesis by both Dall-E and Dreambooth (obtained from internet)
As per our analysis DALL-E from OpenAI has slight edge in realistic image synthesis, while stable diffusion from stablity.ai is a complete open-source implementation.
Comentários