AI characters evolve visually as emotions change during conversation. **How it works:** - 6-axis emotion model - LLM drives conversations with persistent memory (1,000-char rolling context + trauma system) - ComfyUI generates images when emotions peak and situation changes. - Gacha mechanics add unpredictability (watch x5.0 at 0:12) - Dual-core freedom: Use any LLM (Ollama/OpenRouter) + any ComfyUI checkpoint/LoRA (SDXL, QWEN, Z-image) - 100% local option with Ollama