This adapter is designed for Stable Diffusion 1.5 to inject general image compositions into the model while largely ignoring style and content. For example, a portrait of a person raising their hand will generate an image of a completely different person raising their hand. The advantage of this adapter is that it allows for more flexible control, unlike Control Nets, which strictly match the control image. The product was conceived by POM with BANODOCO, trained by ostris, and released by them.