News - Saturday September 10th, 2022

Graphic experiments with Stable Diffusion

In August Stable Diffusion was born, which we can consider a bit the successor of Disco Diffusion.

This new model was immediately implemented in websites such as Nightcafè Studio and Wombo, it is certainly a powerful tool and much faster than its predecessor (from my tests we are in the order of 1/40 of the time it takes to produce an image without the ‘subscription to Colab which allows you to have access to more powerful GPUs).

During the summer holidays I was reading the posts and looking at the results of the first experiments, so I was looking forward to trying this new notebook.

My opinion is that to get better results you definitely need a subscription to Colab (which guarantees longer work sessions as well as faster calculation) and a fair amount of free time available. My initial enthusiasm for these TTI tools is starting to fade, probably due to the immense amount of images posted.
In fact, there are thousands of new users of these AI who call themselves AI-Artists and who have thrown themselves into the world of digital graphics in the last 2 months, especially to sell their works in the form of NFT.

I am very skeptical. I think this thing is getting a little out of hand, but I also believe that everything will normalize just as quickly.

Here are some of my tests.