During an in-depth conversation between Somya and Tippi, Somya revealed her early testing on using Bacalhau for GPU Compute on images was much slower than we wanted (aiming for <15 seconds per image), so we began investigating the feasibility of Optical Character Recognition tech.

Our early experiments revealed it was pretty straightforward, and could read the information about players race and class with high accuracy. Since we have a temporary solution for high speed image generation using Colab (see Dragonborn Backend 🔥) we started diving into the scripting part of making a stable endpoint that can get a screenshot of a fastcharacter.com character sheet and return the ancestry (color) of the Dragonborn (or if it’s not a Dragonborn) and what class the character is. Example: “Red, artificer”