Z-Image-Engineer for ComfyUI: load the Qwen3-4B prompt model (sharded HF safetensors or GGUF) as a Z-Image text encoder, run local in-ComfyUI prompt enhancement with live preview, and run the fast LFM2.5-1.2B Engineer as a standalone enhancer (CUDA/ROCm/CPU) - no external LLM server required.
https://github.com/BennyDaBall930/ComfyUI-Z-Engineer
git clone https://github.com/BennyDaBall930/ComfyUI-Z-Engineer
comfy node install comfyui-z-engineer