Simple worflow With Model and lora loader, that save the metadata with a special node so all the infos you need is save in the image and everything is recognised when you upload an image on CivitAi.
V2 :
New version include SD1.5 and sdxl clip and Img2img. Also reorganized a bit added the lora info node to get the trigger words
Also took out the one button prompt and the ultimate sd upscale.
🔥Include a bunch of my fav nodes 🔥
- https://github.com/rgthree/rgthree-comfy
- https://github.com/pythongosssss/ComfyUI-Custom-Scripts
- https://github.com/Suzie1/ComfyUI_Comfyroll_CustomNodes
- https://github.com/jitcoder/lora-info
- https://github.com/alexopus/ComfyUI-Image-Saver
🧡💛💙
👉Enjoy and Have Fun ! 😎
🧡💛💙
Description
FAQ
Comments (8)
Absolutely perfect for anyone who wants to switch completely to Comfy.
Incredibly good work my friend!
Thank you ♥
🙏🧡
doesn't include lora weight, i'm using a different setup and have the same issue.
i hate that it's on us to do this metadata bs. whats so hard about reading what loras are being used and at what weight?
and you have to manually add what loras are being used? that's too cumbersome
@thighenjoyer , yeah if you change of loras often it's a bit tidious but it works, i found out a node that could add the loras automatically (A Wildcard manager) but i couldnt make it work good till now. i will post updates if i manage to ...
"i hate that it's on us to do this metadata bs."
I'd argue it is not on us. If Civitai wants metadata from my images, they'll make their backend read Comfy's. If they don't, well, it won't make me lose any sleep at night.
On a tangent, sometimes I wonder why they even want to display the metadata. Such a simple listing won't capture complex workflows. If it is there to help others duplicate the process... and if I happen to use ControlNet as posing assist, or IPAdapter to bring in a style, or inpainting to regenerate away some blemishes or sausage fingers, or even switch to another model on upscale... then that data is largely useless and they'd really need to see the entire workflow and any external data that was brought into it.
(Of course, I realize displaying Comfy workflows would place a big burden on their system, and they appear to scrub them away in any case.)
@FeatureEnvy Yeah, absolutely. My main workflow is a monstrosity that literally needs revving up like an old chainsaw to even start generating, no joke, 20+ clicks on the queue button or it won't start when first loaded in. That thing cannot be captured with those few lines of metadata at all. And when I hi-jack my prompt and alter it from pass to pass, plus potential adetailing with its own individual prompt, then some inpainting done with an extra workflow, and when I feel like it add a supir detailing on top of that as well, with addition of tags... yeah.
I use a specific metadata only workflow where I load Checkpoint and LoRA info which is close enough to replicate the root of the actual upload to the point of vague resemblance :D
With all the popularity and touted strengths of ComfyUI, I am also extremely surprised that there is still no convenient solution for managing LoRAs. Metadata for Civitai is one thing, but managing trigger words and selecting LoRAs... that's truly a Sisyphean task.
The solution here is already really good and by far the best I've seen so far (thanks for that!), but it's still a shock when coming from A1111.