Trained this model to assist flux kontext for creating exaggerated proportions. Tagged using natural language with a dataset of 1700 highly rated images from r34. The model is under trained but still works well with flux kontext, and less so with normal flux dev generations.
