How to make multiple character on same image, but keep this level of accuracy and details?

A viral Reddit post from user goku58s has exposed a significant technical hurdle in the AI art community: generating coherent multi-character scenes using specialized anime models in ComfyUI. The creator, who uses Illustrios base models for their anime-style output quality, reported complete failure when attempting to produce images with 2-3 distinct characters, despite experimenting with advanced techniques like regional prompting. This challenge persists across multiple workflows and highlights a fundamental limitation in current stable diffusion architectures when handling complex scene composition with niche model fine-tunes.

The technical difficulty stems from how models like Illustrios are typically trained on single-subject compositions, lacking the contextual understanding for multi-character interactions. Regional prompting—which attempts to assign different prompts to different image regions—often fails with specialized models that weren't optimized for this use case. This limitation affects thousands of anime-focused AI artists who must choose between their preferred art style and scene complexity. The community response suggests this represents a broader need for better multi-subject training data and improved compositional controls in next-generation image models.