AI Art Tools I Use

There are so many tools coming out for AI generated art. What is mind-blowing to me is the speed at which ai platforms are appearing, and the quality on some of these models. They are so realistic. It also raises ethical questions of transparency and usage.

Images are generated using diffusion models. You put in your prompt for the image and 'noise' is generated. Think of snow on a screen on a old TV. The model then slowly works backwards, composing images based on the prompt. It's an intricate mathematical process.

Some platforms have a negative prompt meaning you can indicate items or objects you do not want in the image.

Here are five platforms we have been testing out and using. The magic happens in the prompting.

• Adobe Firefly: We used some of these images for a recent opening video at a conference in Zambia. It was very simple. I like that there's a embedded tag on these photos with content credentials. So people can know looking at metadata that this is an AI generated image. Helps with disinformation.

• Civit.ai: This was extremely realistic. There area lot of great models here that you can use.
• LeiaPix: 2D Images to 3D images. Super cool.
• DALL-E: It's easy to use it on Chat GPT4. The images are cool but they feel less realistic that some of the others I've tried. Perhaps I need to improve my prompting.

• Midjourney: This is on the discord server and it's getting pretty great. I have a subscription here. A new text to video feature is out and I am experimenting.

I've signed up for Pika.Art web based experience, and am testing Leonardo.AI. Trying to figure out how to use Stable Diffusion which is a lot more technical and needs GPUs and computing power. Diffusion Bee is what I'm testing out on my lap top at home.