Stable diffusion video tutorial

How to install (and run) Stable Diffusion

Stable Diffusion (SD) is the most flexible, and open way to generate AI artwork, simply because it can be installed locally - compared to other solutions like artbreeder, or midJourney which are proprietary version - stable diffusion is open for the public.

The hardware requirements though is quite demanding, as you will need a recent computer with a modern AMD or Intel processor, 16 GB  of RAM, a good video card like NVIDIA RTX GPU (8 GB of memory), and a minimum of 15 GB of free storage space available.

You could theoretically install SD  into a virtual machine, but then it will run in a lower speed (not being able to access the video card), generating a simple image of 512x512 will take up to 30mn.

The installer (tested and working as of Dec, 26 2022) is located here (for Linux and Windows) the installation process is pretty long, Python 3 will be installed, then many other python packages... 

Once the download is finished, access the extracted zip file and run ./start.sh (which is the way you will later run the SD as well).

In case you get errors, most of the scenarios are covered on this page

https://github.com/cmdr2/stable-diffusion-ui/wiki/Troubleshooting

As for example, I had this scenario:

-  the installation went fine, and once it was ended, the internet browser popped up with the address localhost:9000 and then this error would occur:

Killed uvicorn server:app --app-dir ... --port 9000 --host 0.0.0.0

The reason I had this error was that my setup for the virtual machine had too little memory (4GB), and unless I setup up to 12GB that error would appear.

First steps with Midjourney

https://www.midjourney.com/showcase/recent/

To get into Midjourney, I followed this nice video from PromptMuse

https://www.youtube.com/watch?v=xdE_h-xjMPs&t=50s

These settings are very important:

and Generated this 3 pages comics for fun on canva.com

https://www.canva.com/design/DAFVkcVC-j8/RlGDZRREcqybgkGryQSeIQ/view?utm_content=DAFVkcVC-j8&utm_campaign=designshare&utm_medium=link&utm_source=homepage_design_menu

First steps with AI learning

 Since AI (Artificial Intelligence) is all over the place right now, I started to look into that topic and found a few interesting websites.

First, AI is really a revolutionary step forward, why am I saying this?

Well, just look at the music, images that are now generated by a simple text prompt, this is insane, this technology is close to magic, really - it is advanced mathematics, but it feels like magic.

You say the word, and the picture appears in seconds, with visual that can be compared to the most talented artist creation. That is magic.

How does it work, in the background, I have tried to get into the basic of 'Stable Diffusion' (SD for short) and basically it comes to Machine learning, there is a very extensive data set of images, or collection of similar photos, that will be the starting point, then a model is made that is the part which is traditionally called the software, the algorithm, thanks to that part SD will be able to produce an artwork that looks like a painting, or in the style of a particular artist, or a visual style.

To summarize:

- Dataset, that will train the machine learning

- model, that will be interfaced with the program to produce a particular 'style' 

and then running the whole thing into a browser

The picture above is 'Black widow' using a style byMarvel Studio, quite decent result

Stable Diffusion

Next level instagram effect with lensa.ai

Remember my post about 'reface, prequel'  filters application back in September 2022?

Well, this is old news, the new filters are now artificial intelligence boosted with lensa.ai (developer, Prisma labs, inc.)

The Lensa AI app is currnetly (Dec, 2022) the No. 1 spot on the iOS App Store's "Photo & Video" charts. "

Lensa AI works by inviting users to upload 10-20 photos of themselves. 

Using the open source Stable Diffusion model, the app processes your photos to generate avatars of you that look like they were created by a digital artist.

• https://www.instagram.com/lensa.ai/?hl=en
• https://apps.apple.com/US/app/id1436732536?mt=8

The number of download for the app has reached 24 millions (Dec, 5 2022)

Z-library down past Nov. 4 2022

Z-library was a database of  e-books and scholarly articles operational from 2009 to November 2022. 

The site was an important resource for many people, most of them college students. 

The website was shut down on Nov. 4 2022 after the U.S. federal government seized its domains.

The domain has been seized, but Z-Library website is still  available via Tor and I2P network.