Since almost all screeshots on the web are trash, had to make all the training images myself using extracted model.
lcstllmn-xl-v2-32-12.safetensors
