A ComfyUI workflow for generating two clearly separated subjects with regional prompt control. The workflow uses separate positive prompt sections for the global image, subject 1, subject 2, interaction, anatomy, scene, and lighting. It also supports global and mask-based regional negative prompts, allowing subject-specific corrections such as preventing gender swaps, duplicate heads, wrong age appearance, merged limbs, and unwanted body-part mixing.The person/subject is always on the white side, depending on which black and white image you download or have created by an AI, and where you use it, you must always describe the white side as a position in the image in the global as well as in the respective person/subject positive prompt.

ComfyUI Regional Two-Subject Prompt Control Workflow

This is just a slightly modified version for a bit more control; I’ve also connected the “Load Image” node to each character's positive prompt.

Files with SHA256: 438d8a379fda2e255e037fe19992dd4cdb4c75404c2168e788ac4858072aaee2