Don't want to miss any of my pictures? Follow my Post Collection

V3

The two v3 versions (with upscale and without upscale) require a custom node collection developed by me. And can be installed with the "Install Missing Custom Nodes" Feature from ComfyUI-Manager. The node collection can also be downloaded via the manager Manuel or via git clone.

https://github.com/BlackVortexAI/ComfyUI-BVortexNodes

How does the V3 version work?

1. Specify the folder in which your images are located

2 (Upscale version only) Define from which resolution images should not be sent to Upscale (the number defines the minimum size)

3. (Upscale version only) Adjust your flux parameters. If desired, the upscale type can also be exchanged. Currently the 1 Step UltimateSDUpscale is configured.

4. Start the run. All processed images are saved in the subfolder “/dataset/[resolution]/with the configured zero_padding and resolution as name, as png and the caption as txt.

V2

It seems that some people have difficulties installing the LogicUtils nodes via missing nodes. If this is the case just search for “LogicUtils” or install directly via url. https://github.com/aria1th/ComfyUI-LogicUtils

This workflow is used to caption images from a folder via Florence2 and sort them according to their min resolution (256/512/756/1024). The images and the caption are saved in the corresponding folder with a consecutive number and the resolution [count]_[res]_00001_.png

The folders can then be used e.g. at https://github.com/kijai/ComfyUI-FluxTrainer

描述:

The two v3 versions (with upscale and without upscale) require a custom node collection developed by me.

Nodes

训练词语: