A crazy node that pragmatically just enhances a given prompt with various descriptions in the hope that the image quality just increase and prompting just gets easier.
This node requires an N-th amount of VRAM based on loaded LLM on top of stable diffusion or flux.
Since ollama keeps a given model loaded via olama run <model> as long the instance is running.
So for example using SD1.5 ~ 5GB VRAM + Mistral:3B uses around 16 GB VRAM (this is just predicted on my system under windows).
Have to be careful when loading too many models since it overloads the vram and you must probably quit ollama and restart the instance.
Ollama must be running or else the node will not be found / visible when starting comfyui.
- Simple node, just enter text and go
- INPUT: CLIP
- FIELD:
- model (this is the model liste loaded from ollama)
- positive text (llm enhanced from selected model)
- negative text (llm enhanced from selected model)
- OUTPUT:
- ENHANCED_PROMPT
- NEGATIVE_PROMPT (atleast what the llm determines, the one selected)
- ORIGINAL_PROMPT
You need to have ollama installed from https://ollama.com before continue.
When within your custom python environment go to the custom_nodes/comfyui-prompt-enhancer folder and install via pip install -r requirements requirements.txt.
If you feeling lucky you can also install via pip install ollama
That should be it.
