[IanCal]

This leads to the more interesting side that he's built a custom inference setup to run more efficiently.

> TextSynth employs a custom inference framework to get faster inference (hence lower costs) on standard GPUs and CPUs.