Running the Studio's engine on your own machine or a free Colab

Hathor Studio publishes the exact engine it runs, as one Python file, so anyone can run the same pipeline on their own capacity. The instructions are on the Studio's engines and Make it yourself pages: https://hathor.soapbox.community/engines Why bother - The Studio's free server renders one image at a time, and when several people are waiting it tells you and offers your own engine instead. - On a GPU the same pipeline is far faster than the server's CPU. The three routes 1. Your own computer. Works on CPU (slow, like the server) or an NVIDIA GPU (fast). 2. A free Google Colab notebook. Colab's free tier gives a GPU when available, with session time limits; you lose the session when the tab sits idle. 3. Your own Modal GPU, about two seconds an image according to the Studio's page. Connecting it After you start your worker, connect it from the engines page ("My own worker"). The Studio then sends your requests to your engine. Keys - Keys typed into the page stay in that browser and go only to the chosen provider. - Keys kept in your Pentecaust account (Pentecaust Connect) are encrypted at rest, and the Studio receives only a short-lived pass that can make images; it never sees the keys. See https://wiki.soapbox.community/wiki/Pentecaust Report back here Post your hardware and your seconds per image. A table of real numbers from real machines is exactly what this board is for.