AI 藝術與 Google Colab 第 1 部分

This class will introduce students to AI Art and some basic strategies to make better prompts (the request for an image from the AI).  These will include settings, modifiers, and artists styles.  We will also make use of negative prompts to prevent distortion of figures.

We will be doing our work using Google Colab which means our art will be created by our code, without any websites as intermediaries, simply by temporarily downloading files to our workspace directly from Stability.ai (the company that owns stable diffusion).  The files downloaded contain built-in filters to black out inappropriate images.This will give students a tremendous amount of freedom going forward as new models and tools are released.  This also allows students to make many versions of an idea and choose the best one.  

After a short intro to AI art we will step through the notebook and see how each of our strategies improves the image of a Boston Terrier in Mecha Battle Armor.  We wil also see how we can make this image in the style of many artists.  After this we will start tweaking the Notebook to change our image topic (prompt) and the number of images produced.

Finally we will give a teaser for ControlNet.  The image of the robot painting a self-portrait (right side of class image) was created from the original (on the left side).  ControlNet allows us to create an image using the structure of another.  This is just one direction in which AI art is going.  Not only can we create images form prompts but also form other images or even just poses or all combined.  What happens is we usually end up with something that was better than we could have imagined.  The technology is incredible and I am delighted to share this with you.