*1/19/24 - ive uploaded it as converted safetensors file. Maybe that will work. If not ill figure out another way.
An experiment in merging positive and negative embeddings. Came up with a fun one that creates mostly a faded inked portrait and sometimes a scene. I like it. Others have liked it, somewhat, or they are just trying to get Buzz by liking it, who knows. The beauty of the buzz system is that it makes me think that people like it. This makes me feel creative. So, I create more. Which makes me happy. Go Team.
The negative embedding for the PopSkull model is listed here as well.
sincerely, DjNastyMagic
Description
this all came from merging my own 0 training embeddings that were created with the embedding inspector extension in automatic1111. lot of merging in order to get the desired effect but I've used it on many of the pictures that I've uploaded and those seemed to get a decent amount of likes. you may see "Chode***" in the sample images because that was the original name of what the PopSkull model was going to be called an acronym for Collected Hauntings of Divergent Expressions. ;)
FAQ
Comments (14)
Sorry, this doesn't show up for me?
What doesn’t show up where?
very helpful.
@oskarsson At the time that you posted this no one else had said they had a problem with it, so i was asking for more clarification because “this doesn’t show up for me” isn’t specific enough for me to know where the problem is.
! apologiez for being salty. It doesn't show up as an embedding or a Lora, so I'm not able to test it or use it. Too bad, the preview images looks great!
Do you know if anybody has got this to work?
Edited: I checked out your models and this seems to be your first embedding. I'm just saying.
@oskarsson thank you. ive tried converting it to a safetensors file. let me know if that works. The issue is that i have no problem running it in automatic1111 or via the Draw Things iOS app and those are the only two i use.
"This file may be malicious, so the program is not going to read it."
You are failing safe unpickling. What do you have in there, hmm hmm hmm?
that's odd, the only thing I did was rename the file from Chode75v5r to WSOTA-Dark. I honestly haven't tried it since I renamed the file because I didn't think that mattered. ill upload it as I originally created it and see if that fixes it.
ive uploaded it as originally created, too much work to spin up automatic1111 just to rename something. assuming its a header mismatch because of the renaming raises a flag.
@djnastymagic merging embeddings can lead to this type of issues. Many embeddings are trained with different kind of trainers (some from the in-built AUTO1111 system and others from Kohya and whatever) and their metadata might get messed up during the merge.
I suppose that you used the Embedding Merge extension for AUTO1111, as far as I know you can EVEN merge embeddings from SD2.X (with usually bad results) for to have even more chaotic effects.
@gsgsdg Thanks for the info. Yes I did use Embedding Merge as well as Embedding Inspector to create this. Any idea on how to fix it? Why does it work for me but everyone else is getting errors?
I got the same error for some reason. "The file may be malicious, so the program is not going to read it.". I ran it through picklescan locally and I get nothing so I don't know what the problem is.
I'm awaiting the uploaders response because I got the same warning from my 3+ virus scanners.
@SOLTIS I’ve re-uped it as a safetensors file. It should work you may just need to rename it. It’s showing as an error because merging the Embeddings removed the metadata and/or it’s looking at the headers in the file to see what checkpoint or amount of steps it was trained on and there are none bc it’s not a trained embedding. In the future I’ll make sure to do this correctly. But I promise it’s not malicious. It’s basically a glorified prompt.
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.