This workflow requires my Comfyui-lmstudio-prompt-enhancer custom node which is installable via Comfy Manager .
Learn more about this custom node
🛈 IMPORTANT!!!
You will need to search for, download, and load the model llama-3.2-3b-imagegen in LM studio. You will also need to start the API server and generate an API Key.
Learn how to configure LM Studio here.
Notes for using llama-3.2-3b-imagegen
This model has been fined tuned for SDXL and Flux cinematic prompt generation.
This model is small only 3 billion parameters so expect it to be a little bit stupid when it comes to understanding instructions, expect there to be many failed attempts, just increment the cache _buster and try again , when this LLM does understand the instruction it does a very good job.
Keep max tokens 80, the max token length for SDXL prompt is 77 tokens but only 75 tokens are usable.
I changed the temperature to 0.5 you can experiment with values between 0.5 and 0.7
This model likes to use the words "frame" and "framed" a lot, so it is good to include "picture frame" in your negative prompt.