Printing portraits

Each case in the gender violence installation now has a face. Instead of generating profiles at run time on the Pi, I now generate all the profiles I need ahead of time. These are based on this study. Next I use ComfyUI with an open-source model to synthesize fitting portraits based on attributes such as name and age.

I was surprised by how much stronger an emotional reaction the portrait triggers.

Next up:

• Some fine tuning to improve the prints (better portrait-to-attributes matching, more variation in portraits, better dithering, different font).
• Use the HSPOS kiosk printer with the Pi to build the physical installation.
• See whether auto-cutting the paper after each profile is a good idea.