Reminder:
The older embeddings do not work on SDXL or Pony models, they give you errors on external image generators. Not sure what the internal generator does but I think it just ignores them if they throw errors. In the other direction, it works the same, no sd1.5 embeddings on SDXL or Pony and no SDXL/Pony embeddings on SD1.5...
Set of TIs that gives me lots of detail in the skin and hair or even overall detail on the whole image. All but OverallDetail shifts the focus of the image to parts that get more detail now.
Check the comparison on the OverallDetails page but keep in mind that those are done without any other negative prompts or helpers unless stated.
The SkinDetail and SkinHairDetail sometimes have a side effect that shows scars or otherwise broken skin, that can be worked against with adding "skin markings" into the negative prompt.
The XL / Pony versions are tested with the XL Model Fenris (see the resources) and also my own model blend between a pony model and Fenris.
Description
Adds more detail to the eye like a shine and spark, also whitens the eye ball a bit. Works independent of eye colors. Might reduce slight bleeding of eye color. Without any prompt on the eye, it produces unnatural glowing eyes, be aware.
FAQ
Comments (28)
How come this has one star even though it has no ratings and seventy-four likes?
Can 'face detail' do faces in different poses, or just looking forward?
maybe it works, maybe it doesn't. none of the showcase images really show off anything worth downloading imo? and it's not clear without clearly labeled versions with and without the TI
This is great and I use it on many of my generations. I did start noticing that denim was creeping into my results and couldn't figure it out. Then I noticed that they were overalls. It might be good to change the trigger word to something that doesn't contain the name to a common piece of clothing.
very useful
I don't know how to use it. Should I use in Positive Prompt or Negative?
hi, with the eyes and hands and over all detail versions, where do I put the prompt do I put them in the positive or negative prompt? please help would be appreciated ;-;
I'm not 100% sure but it seems to provide some results when used as a positive. But yes, some instructions would help
How to install please help
i know stupid idea but how use embeding
Is there a chance for a sdxl version?
Can you use this on the pony diffusion? I NEEEEEED THIS
Can't wait to use this in SDXL!!
Is it possible to add this to on-site generator?
Very helpful lora
Might want to change the trigger word, keeps putting people in dungarees and overalls! 😁
i like this!
After getting a comment that the new skin detail embedding did not quite do its job, I tested them and found out that pony based models need a higher weight on their embeddings. Check the images for the skin detail embedding, the SDXL model has an extreme reaction to the embedding and the pony models show almost no change. So, if you use this on a pony model, play around with the weights, a weight of 2 sounds otherwise extreme but in this case it might give you a better effect. I tested weights up to 5 and at 3 it starts to break and give you bad images.
It looks like Pony models do not get much effects with the embeddings as SDXL models do regardless of weight, I even tried to remove the score tags and the source tags but same difference...
Embeddings like these do not seem to work on pony models, sorry... I will keep them up as they work nicely on SDXL models...
I tested this a bit further, all my embeddings are text based and if I just type in the text in the prompt instead of the embedding, I get the same issue, heavy effects on SDXL models and almost no effect on Pony models. The way I use these embeddings is to make the prompt easier to read in the sense that it is shortened and I could write novels in the embedding and still have a short and precise prompt to use.
overall good, need pony ver
how to change the trigger word? i keep getting characters with overalls
Is it necessary to use both trigger words, 'detailxl' and 'OverallDetailXL'?
For some reason I can't add it
Just like Cathy Powers it doesn't appear in my embedding list (ComfyUI), however it appears in the lora list if I put it there.
Is this negative embedding, why authors cant tell us this small simple little fact?Also where the hell is OverallDetails supposed page with comparisons
you can use these for PonyXL?
great work, it would have been more helpful if you put each model's pros and cons in the description
If you use modern UI without 77 token limit (all modern UI solved it), you can put normal tags instead of "embeddings", it will work exactly the same.
Only point of using pre made embeddings is overcome token limit and put more tags in prompt (It was a thing in old SD 1.5 with primitive tokenizer which trimmed all after 75+2 service tokens).
Also if you use pre made embeddings, you have no control over full prompt weight, and embedding can ruin prompt weights balance - You can see ruined balance on demo image "yellow eyes.
(Just a little info about pros and cons)
Still one of my favorite, when it comes to realistic images.