So the next thing is a text-to-pose model, so that you can describe your pose and your image with separate prompts, and have them combined via controlnet.