ZEMS – AI-driven emotional evolution with visual character changes

Summary: ZEMS is a local AI system that evolves character visuals based on a 6-axis emotion model during conversations. It integrates large language models with persistent memory and ComfyUI image generation to reflect emotional peaks and situational changes in real time.

What it does

ZEMS uses an LLM with a 1,000-character rolling context and trauma system to drive conversations while tracking six emotional axes. ComfyUI generates images when emotions peak or situations shift, with gacha mechanics adding variability. It supports any LLM (Ollama/OpenRouter) and ComfyUI checkpoint/LoRA models, running fully locally if desired.

Who it's for

It is designed for users seeking AI characters that visually evolve with emotional context in conversations, particularly those using Windows with ComfyUI and Ollama or OpenRouter APIs.

Why it matters

ZEMS addresses the challenge of dynamically reflecting AI character emotions visually and contextually, enhancing interaction realism through persistent memory and adaptive image generation.