Complete flow of ComfyUI integrated text generation
To generate an image using Nunchaku in ComfyUI, you need to follow theFive-step workflow::
- environmental preparation::
Install the ComfyUI-nunchaku plugin (via Custom Nodes Manager or comfy-cli) and download the FLUX.1 model files to the models directory - Loading workflows::
Import the example json file (e.g. nunchaku-flux.1-dev-pulid.json) and the system will automatically configure the node connection - parameter optimization::
in the Nunchaku Text Encoder Loader node:- Enable the use_4bit_t5 option to conserve video memory.
- Setting t5_min_length ≥ 512 improves quality
- Adjust general parameters such as seed, steps, etc.
- input prompt::
Enter the English description (e.g. "cyberpunk street with neon lights") in the CLIP Text Encode node. - Execution generation::
Click Queue Prompt and it takes about 30 seconds to complete the generation on low end graphics cards
Advanced Tips:Combined with the PuLID v0.9.1 module for more precise style control, the editing intensity can be adjusted via the timestep parameter.
This answer comes from the articleNunchaku: an inference tool for efficiently running FLUX.1 and SANA 4-bit quantization modelsThe
May not be reproduced without permission:AI productivity tools " How to use Nunchaku for text-to-image generation via ComfyUI?
