SD Tune - Stable Diffusion Tune Workflow for ComfyUI

A ComfyUI workflow for the Stable Diffusion ecosystem inspired by Midjourney Tune. Quickly generate 16 images with SDXL Lightning in different styles. Load your own wildcards into the Dynamic Prompting engine to make your own styles combinations. The SDTune engine makes 16 images in one batch one after the other thanks to an AnimateDiff node meant to control frame generation.

Select one or multiple images in the style you like, and then put your prompt in the High Quality generation workflow. When you are ready, hit "Progress Images" and the image (or combination of images) will be sent to generate at a higher resolution / quality with your tuned style.

Once you have a style you like, generate ultra high resolution upscaled images with an Ultimate Upscaler node.

Installation notes:

Make sure to load in wildcards to a wildcards directory in your ComfyUI base folder.

The prompt engine can be finicky with punctuation and symbols in certain wildcard sets.

Make sure all nodes and comfy are up to date.

Make sure to update all models and paths used in nodes before use.

Custom nodes for math, image choice, dynamic prompting, ip adapter, etc will need to be installed.

v1b Notes - Changed int node to primitive to reduce errors on some systems. Added clip vision prep node.

v2 Notes - Switched to SDXL Lightning for higher quality tune images, faster generations and upscaling. Added some image enhancement nodes. Implemented IP Adapter v2 nodes.

Thanks to -

RunDiffusion

Chris Goringe for the awesome image selector node!

Matt3o for IP Adapter.

Countless other node creators :)

Everyone who tested!