Atomix Flux Unet txt2img workflow

There are two sub-versions:

Using Self-model to upscale images: using Flux as the upscale model, with higher output quality, better facial and hand details, but the speed is slower.

Using SDXL to upscale images: using SDXL as the upscale model, with low configuration requirements and faster speed, but with lower output quality.

Feature

• The image generation data contain the using model,Lora,embedding information.

• Calling lora in prompt words or nodes（but the lora called from prompt words will not show on civitai web page）

• Automatically adjust the size of the upscale tile and provide multiple aspect ratio presets.

• Integrated with UltimateUpscale and able to switch different models for upscaling.

• Integrated with FaceDetailer and able to switch different models for FaceDetailer.

• Showing FaceDetailer before/after comparison.

How To Use

1.This is a ComfyUI workflow,only for ComfyUI,so make sure that you ComfyUI is available. https://github.com/comfyanonymous/ComfyUI

2.Drag the JSON or PNG files from the zip file to the ComfyUI interface. If everything goes well, the workflow is ready to go.

3.In the step2, there may be a alert for missing nodes, and you need to install these nodes. If you do not want to manually install them, there is a ComfyUI plugin that you must have: ComfyUI Manager https://github.com/ltdrdata/ComfyUI-Manager After installing this plugin, there is an automatic installation function for missing nodes in its management interface.

Related models

CLIP : comfyanonymous/flux_text_encoders at main

VAE : ae.safetensors · black-forest-labs/FLUX.1-schnell at main