"Fixes" A? :shrug:

It makes babies.

Initial D refference

lack of instructions is always a problem,

but embedding is embedding, regardless of dirty pickle or securer safetensor

@asdfgyxcv then why doesn't it appear?

@lnomsim that, i do not know,

i know neither, the technical aspects behind it, nor, you're machine,

all that i can tell you is, that i can see it within comfyUI - with link to civitais page and everything ...

If you're using ComfyUI, embeddings don't appear anywhere, you use them by putting the trigger word in your negative prompt. Or you can install the custom node "Embedding picker" that will list them and output a string that you need to concatenate with your negative prompt

@sanglyon I am not using comfyUI, actually, when I launch stable diffusion, I get the following error:

*** Error loading embedding negativeXL_D.safetensors

Traceback (most recent call last):

File "C:\stable-diffusion-webui\modules\textual_inversion\textual_inversion.py", line 227, in load_from_dir

self.load_from_file(fullfn, fn)

File "C:\stable-diffusion-webui\modules\textual_inversion\textual_inversion.py", line 192, in load_from_file

assert len(data.keys()) == 1, 'embedding file has multiple terms in it'

AssertionError: embedding file has multiple terms in it

I have the same problem. It's not showing in A1111 if I place it in embeddings.

@petardapocpocc I redownloaded the file yesterday and it finally worked.

But I also updated webui to 1.6 version, so I don't know what did the trick. Hope it helps.

@lnomsim No, you need to place the file in: "C:\stable-diffusion-webui\embeddings"

doesn't work for me either, put in folder: embeddings, but cannot find in in webui.

did you guys remember to refresh? it's not auto-detected when a new file is added.

https://stable-diffusion-art.com/embedding/

looks loke it is an embedding - not a LORA

@OliviaRossi I agree but the file extension is .safetensors that is why I am confused

Place it in "webui\embeddings"

Embeddings just has to do with text. Think of a paragraph condensed into one word

Even though it has been a while, I'll answer for the people that still don't know. Probably because the two answers given, don't explain what this embedding does.

The short answer: No one actually knows, except for the creator and even then it's questionable.
Embeddings are similar to Loras, only that they don't add the dataset training to the "knowledge" of the model/lora. The embedding uses the knowledge of the model, in combination with your description and the dataset to create a single "tag/keyword" that combines tokens that can describe your dataset/subject/style.
This also means this single tag/keyword uses up multiple tokens of your prompt.

In this specific case, what we probably can guess, is that the creator used a lot of unwanted things for his dataset and combines this with typical negative prompt tags. As a result you get an embedding tag that sums up things you don't want. But what exactly this is, depends on the creator. Also in a lot of situations, it's not even used correctly, to a point that it's a placebo. People don't know and just add it, thinking it will do something. The best example is the fact, that old SD 1.5 embeddings don't work with SDXL/Pony models, but people put it in their negative prompts all the time.

It can also be seen as bad quality work, to use negative embeddings, because they can be extremely style altering. Showcasing your model/lora quality with negative embeddings in every single prompt, says a lot about your work.

@vynenave926 Extremely well-put, A+