AI Image Generation - Week 9 (tri 2)

Julia Toczyska
Apr 9, 2024
2 min read

Updated: Apr 20, 2024

Today, we began talking about Ai generated art that we find inspiring and/or interesting. I was paired with the best woman, Cassie, to discussing it in pairs! We used Lexica, a website great for exploring community made Ai art.

Scavenger Hunt

Generative Art

Generative art is art that has been in whole or part created using an autonomous system, eg. computer algorithm. The artist may have some degree of control over the algorithm, however, they do not create the artwork themselves.

AI Gen Artists

Surprisingly, there is a lot of artists out there who already use Ai image generation as part of their work! Out of all the artists we had a look at today, I found Beth Frey's work the most fascinating. She produces highly disturbing images and videos which are completely nightmare fuelled... I kinda like it! Sadly, I couldn't find which Ai tools she uses specifically.

Beth Frey @ Sentient Muppet Factory

Types of image generation

Text-to-image
Image to image
Inpainting
Outpainting
Custom models

What are the current best image-gen tools?

Midjourney creates images with incredible detail. It offers high customisation, allowing you to refine and guide the image generation process. HOWEVER, Midjourney primarily operates within Discord which can be confusing.

DALL-E excels at generating photorealistic images that can be strikingly close to real-life photographs. It's ability to edit and make in-depth alterations to generated images offers a high level of control. HOWEVER, DALL-E can sometimes be overly literal with text prompts, potentially missing nuances or creative directions.

Stable Diffusion's massive community creates a wealth of resources, support, and custom extensions. This gives Stable Diffusion incredible flexibility, as developers can adapt and customize it for specific uses. HOWEVER, setting it up, especially with custom scripts, can be annoying. Additionally, it can require powerful GPUs.

Adobe Firefly is still a relatively new tool and might not be widely accessible. Additionally, it is likely requires an Adobe Creative Cloud subscription.