We’re testing out a new custom model trained on 300MB of fiction stories. This is best used in STORY mode (not DO or SAY)