Extensions/ComfyUI-llama-cpp-prompt-enhancer
ComfyUI Extension

ComfyUI-llama-cpp-prompt-enhancer

Local prompt enhancement for ComfyUI powered by llama.cpp, run LLM models natively to refine & generate image prompts.

By wanaigc·Created a day ago·Updated a day ago· 1
wanaigc/ComfyUI-llama-cpp-prompt-enhancer
Nodes—
On cloudLocal install
Stars1
Updateda day ago
Readme

ComfyUI-llama-cpp-prompt-enhancer

Local prompt enhancement for ComfyUI powered by llama.cpp, run LLM models natively to refine & generate image prompts.
[📃中文版]

Preview

Features

  • Local prompt enhancement — Refine & expand your image prompts with local LLM/VLM models (llama.cpp based): fully offline, no API keys.
  • Multiple prompt styles — Built-in presets: Tags, Simple, Detailed, Extreme Detailed, Cinematic, plus a Creative Refine & Expand mode.
  • Model-specific prompt skills — Prebuilt skills for Minimax H3, Krea 2, Moss Image, Qwen Image 2.1 to craft prompts optimized for each model.
  • VLM image-to-prompt — Load vision models with mmproj weights (incl. Qwen mode) to generate prompts directly from images.
  • Preset system — Save, manage and reuse your own prompt presets (Preset Manager / Save Prompt Preset).
  • Structured output tools — Parse JSON from model responses, convert to BBoxes / SEGS / MASK for grounding workflows, and unpack markdown code blocks.
  • Any model path — Load .gguf models & mmproj from any absolute path (not limited to ComfyUI/models/LLM); connect directly from the Wanaigc File Finder.
  • Model management nodes — Loader, Parameters, Unload Model, Clean States, and a BETA Text Encoder node for fine-grained control.

Installation

Install the node:

cd ComfyUI/custom_nodes
git clone https://github.com/wanaigc/ComfyUI-llama-cpp-prompt-enhancer.git
python -m pip install -r ComfyUI-llama-cpp-prompt-enhancer/requirements.txt
#For CUDA 13.1, please run:
#python -m pip install -r ComfyUI-llama-cpp-prompt-enhancer/requirements_cu131.txt

Download models:

Credits