A pretty simple workflow with a few common features.
I had to do a downgrade of my node Save Image because I spent several days battling with the Comfyui registry scanner and couldn't find another solution. If a prompt is sent to KSampler via any transformation like my anyconcan node, but it won’t extract or save metadata. The solution: manually enter positive/negative in Save Image.
Description
v4.7:
Replaced the node that loads images on my own node.
Added a metadata display node.
Changed the default loaded model for Florence-2 captions.
Updated the readme file.
v4.8:
Minor fixes.
FAQ
Comments (6)
Hope you update your SDXL one, i still use it
@slln , I don't have any ideas or tools for improvements right now. Actually, I spend most of my time battling with ComfyUI’s frontend, which has gotten progressively worse—especially the 'legacy mode.' This takes away from my development and research time. However, the workflow for SDXL-like models should work fine. If you encounter any issues or bugs, let me know, and I will try to fix them.
@DemonAlone I will! yeah, the nodes2.0 stuff slows my (very beefy) system to a crawl. such is the way with open source stuff thats angling to get bought out
TBH it was mostly a "don't forget about the SDXL stuff, theres still a lot to like and enjoy about pony/illustrious models, both style and speed-wise"
@slln , Nodes 2.0 are not only about speed. That 'parallax' effect—when links and nodes separate while moving the canvas—makes me feel worse than the freezes in VR. Newer models are cool, but they have their own limitations. I would abandon sdxl-like models only if I found one that was better in all aspects (or at least most) than the sdxl-like ones. I don't think this will ever happen, so I keep working on my workflow. My StyleSelector node/extension was part of that work and consumed too much of my time and energy.
@DemonAlone heard and agree. a bunch of weird UX decisions for sure.