This is LITERALLY just so you don't have to keep typing all of your scores in when you're doing generations. The images are just showing use of it, there should be no real need of an XYZ plot.
There's no other modifiers in it, it's created using the SDXL branch of Embedding merge.
https://github.com/klimaleksus/stable-diffusion-webui-embedding-merge
Designed by Ktiseos Nyx.
We've been at this since January of 2023! If you're interested in supporting us in our long standing AI journey please scroll down for ways to connect & sponsor us! Currently if you're interested in having it on the generator please feel free to donate towards auction costs!
✨ Recommended Settings
To get the best out of ANY of our models, try these settings:
Samplers: Euler or Euler Ancestral *(SDXL) - Or your preferred DPM style for SD 1.5 - We don't have settings for Flux as we don't gen flux off Civitai just yet!
Scheduler: Karras
Negative Prompt: Generally not needed! The model is designed to perform well without one, but feel free to experiment based on your preferences.
HiRes Fix:
Steps: 15
Denoise: 0.3
Upscale: 2x
Upscalers: Rex, 4x_Remacri, Lollypop or your personal favorite.
ADetailer: Use sparingly. Due to potential VAE differences, it might occasionally produce grey squares.
Embeddings: We've used some of our own custom embeddings to showcase character types in our examples, but no stylistic embeddings were necessary to achieve the core look.
Checkpoint Compatibility: Depends, not everything will behave.
🛠️ Training Details
If you're curious about what's the training details in full make sure to use this tool: https://xypher7.github.io/lora-metadata-viewer/
🙏 Acknowledgements
A huge thank you to the incredible community members whose work and inspiration contribute so much. Special mentions to:
Arthemy, Wai, Holostrawberry, FallenIncursio, RichyRich, JustTNP, Hoseki, CitronLegacy, Samael1976, Kairen92, ShadowxArt, Flyx3, YeiYeiart, LazyVisionary, and so many more! We try to train as much as we can ourselves, but the community's creativity is a constant source of inspiration.
🌟 About Ktiseos Nyx & Connect With Us!
Ktiseos Nyx is available for commissions! Find us and more of our creations here:
Discord: https://discord.gg/HhBSvM9gBY
Our Other Socials: https://beacons.ai/duskfallcrew
Sponsor Models & Commissions: https://ko-fi.com/duskfallcrew/commissions
TensorArt Space: https://tensor.art/u/611011406535381539
Dataset Tools: https://github.com/Ktiseos-Nyx/Dataset-Tools
💵 Training Costs Money!
While Civitai style trains usually cost buzz, testing and infrastructure to create better quality content to service our customers long term does cost money.
We're not looking to be the next Civitai at all, but AI has been a format of income outside of the inability to find regular work.
Donations, Commissions and Sponsorships are the backbone of our work!
Buzz donations work to keep our models on the generator that YOU the user choose! (This was originally a rarity, buy recently you've all chosen and we're all proud!)
Lora training can take time and effort, and while we do have cost layouts - if you're wanting a simple character lora you can use the Ko-Fi link shown in our link In Bio Above!
✨ Commission Details
If you're wondering why you're seeing a style lora or other loras ahead of your queue'd training or commission please see these FAQ lines:
Style Loras are usually developed independent of our normal training queues at the moment, as we use Civitai's platform to train them for use with our normal model making schedule during the month.
Character loras normally are trained as they come in, these semi depend on our own training service we've developed. Right now the NextJS system is almost complete for testing, and we will be resuming PAID and Free queues ASAP.
Concepts that are CLOSE to "STYLES" are again created using Civitai, at least on the short term. Once our main service has been completed we will amp up our schedule.
Description
Same thing just a safetensors file
FAQ
Comments (32)
Such a good idea. I wondered if there was some reason why it couldn't be done. Part of what we're asked to type are some scores in the negatives. Is that covered by this?
ARE WE? LOL Please let me know because All i went by was the positive scores -- IT LEGIT covers most stuff for me
@duskfallcrew Ohh I looked it up, and we aren't. It's just super common practice, and some mixes say they require it (putting "score_4, score_5, score_6" in the negatives). But yeah, it might be one of those superstition things. I'll definitely use this. Thanks!
@duskfallcrew No for pony you are suppose to use the entire string: score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up, What is score_9 and how to use it in Pony Diffusion | Civitai
TLDR is that the model learned unintentionally that the whole string = good images
@bran339 CORRECT, this is an embed made with EMBEDDING MERGE which takes strings of text and splats them as vectors to tnker at the text encoder (dont ask me the full details i just make the toys XD) - IT LITERALLY HAS "score_9, score_8_up, score_7_up, score_6_up, score_5_up, score_4_up," in there - that's LITERALLY it there's no source in it, there's no other things it's just the score string :) - it's so i don't have to type it in every single time :P
Also: not everyone uses the FULL one, but the full string is indeed in there because of Astralite's intentions with the model.
Also trick I Learned:
I've trained a few lora with adding the SCORES to the tag set? It actually helps retain styles XD
Ok YES we do provide a negative score now, https://civitai.com/models/389486?modelVersionId=569007
Double tapping the praise! The convenience really speaks for itself. :)
Convenient!
Is there an easy way I can convert this to safetensors format? SDnext wont load pts when using the diffusers backend.
Oh lord - i'll see if i can find. a way to do that this weekend for you!
You still have to put the bad scores (score_4, score_5, score_6,) with this one?
..... those are bad scores?
Since when?
All scores are built into this one ACCORDING to how I was taught how pony XL prompting works. and that includes score 4 etc -- unless you're meaning score_4 without the "UP" - this is the positive prompt one, i'm STILL trying to parse what negatives i've made that i'm missing :3
@duskfallcrew I was taught that you should put "score_9, score_8_up, score_7_up, score_6_up," in the positive, and "score_4, score_5, score_6", in the negative. That's how I've been doing my images. Am I doing it wrong? Now I'm confused.
@MacabreGinger HMM No - Nobody's WRONG, just that I half the time don't pay as mcuh attention to how pony works XD However, i use Embedding merge, so i'll consider making that an embed soon!
@MacabreGinger Just because you instructed me as a wise fellow pony user! I made the embed using those scores! https://civitai.com/models/389486?modelVersionId=471883 I'm an idiot and i dont know how well it works but i used it like 100 times last night
@duskfallcrew Me, wise?
@MacabreGinger Wise enough to teach me something new! :P
@duskfallcrew So let's be clear here. does it work? Are the images better with the negative embedding, then?
@MacabreGinger Not really, iv'e never actualy before you posted that used that - i made it because ther'ell be people that STAND directly by it - it'll largely DEPEND on what you're using it for and where it's for - those scores depend on certain styles sometimes i think
@MacabreGinger As for the score embed that goes in the positive, that definatley works -xD it eliminates needing to type it all the time - but the use of it depends largely again on what you're looking to do - I use it a lot but there's been SOME content i drop it for
Okay, so does this embed only cover the positive "score" prompts, and not the negatives?
Can confirm that it works with minimal changing to outputs.
Results from one image generated without the embed, using the regular score stuff, and an image generated with the embed instead are nearly identical. There's some very slight differences, but I could just as well chalk that up to the difference in token count or Xformers being weird.
XD That's good because it's literally the scores prompt in an embed :P
@duskfallcrew Obviously, but whether or not it does what it says on the tin is what's important. All your showcase images are just random images, there's no comparison shots of using the embed vs regular prompting. Especially since you have to train a TI, some variance might crop up due to training data or it might have a hard baked style. Unless there's a way to just put prompts solely as text and save it as a TI. TI's have been under my radar for a while, so that might be the case.
@anon4324 Yea there is LOL, it's called Embedding Merge :P - I don't train my embeds, it's a nightmare. And variance crops up EVEN WITH embedding merge because the text gets convered to the vectors that a textual inversion do.. 'll link you with the repo in a tick here:
@anon4324 As promised: https://github.com/klimaleksus/stable-diffusion-webui-embedding-merge
@duskfallcrew Woah, really neat. Gonna have a go with this one, thanks.
It suggests setting clip skip to 1, but most if not all pony models use clip skip 2.
XD so use clipskip 2 then :) <3 it's just a suggestion xD It's a default setting
how to change the clip skip on pony??, im using sdwebui forge and cant see the slider.
@fronyax there's some boxes like SD, FLUX in the top left corner. Check "all" then clip skip will appear.
@fronyax did you ever solve this? You not at my setup right now but there is a setting you need to change for it to show in the ui.. google it and hit me up if you can't find it.
thanks
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.