This is a textual inversion embedding for D.va (Hana Song) from Overwatch.
Use corneo_dva to activate. It's a little overfit so you might want to de-emphasise it in some situations to (corneo_dva:0.9) or (corneo_dva:0.95). It also works best below about 11 CFG value.
Example images produced with OrangeCocoa available here https://civarchive.com/models/4355/orangecocoa-5050-mix
Check out my other civitai models for more characters and sex positions! https://civarchive.com/user/Master_Corneo
If you enjoy this embed, I genuinely love seeing your creations with it - you can attach them to a review if you're not aware (but not comments sadly)!
Description
BIG UPDATE! Turns out an extension I was using as part of my embedding creation just, er, doesn't work most of the time, so all of my training data and embeddings so far have been a bit wrong.
Put this theory to the test by re-training D.va here and it's right, the embedding is much more accurate now, especially things like the eyes didn't really work correctly before and you had to tag brown_eyes and they still looked bad. Now you should get somewhat decent eyes by default!
FAQ
Comments (19)
Here's my hope a Mei TI comes soon
I think I have all the overwatch girls on my wishlist of requests at this point :D
Wow, may you mind sharing what was wrong :)? This embeed quality looks a lot better! (Btw i am having problems with eyes too)
Sure, so what I was doing was using the Tagger extension in the Web UI, specifically the "Exclude tags" part to remove the tags that I wanted to train towards. Turns out that only works about 30% of the time, and most of my tag text files were still including the tags I thought I had removed, and in short it was wrong but it kind of brute forced itself into a working TI with a long training time, but issues remained.
Now I'm using the Dataset Tag Editor to remove the tags I want to remove, and it is SO much better than what I was doing before, never mind the fact that it actually works this time.
So good news is all my future stuff should be better (and take half the time to train!), bad news is I feel like I have to fix all the currently released ones now.
I think im going to make a copy of the automatic1111 with only the novelai model,no Vaes,and only the training extensions because i think something is fucking the training..
That's a good idea I think
Great job on these!
Think you can make Moira from Overwatch? :)
Unfortunately there just isn't anywhere near enough art of Moira to make one, especially not enough that's consistent. Just nowhere near as popular as the standard Overwatch waifus sadly
@Master_Corneo That's what I was afraid of :'(
need assistance. ive installed this embedding. when i generate any image using the prompt. i get a result image that doesnt contain this inversion.
i am certain the embedding is loaded as i am getting this line in the SD: Used embeddings: corneo_dva [b1d5]
ive tried using prompts from uploads i found on the site and the same issue is encountered
I have the same issue and tried everything, but it doesn't work. I think the file is corrupted or I am doing wrong. Did your problem solve Please tell me
I can't tell you entirely, how much i appreciate you. Thanks to you, I can generate my D.Va.
Please, Please make Lora of Echo and Orisa. Please.
wow the porn is so popular, not only does it have 2 games a game, but LORAs too?
Hi thanks for sharing! I am super new to SD. what sampling method did you use?
Welcome! You can see that in the image metadata here. I just looked and have already forgotten by the time I came to write this response!
Master Corneo, welcome back! Could you please update this embedding to xl version? So far i was getting the best results with this one on 1.5 but..would love to move forward :)
Thanks! Now I use Pony (or actually Autism Mix) as an XL model as it has good knowledge of popular characters like D.va built in, so I don't think there's a need for a character embedding like this unless it was for an alternate outfit or something
@Master_Corneo Ohhh i see. Yea, i'll try this model then
Please help. Whenever I try to download this model, I get redirected to a page that says:
{
“code”: “not_found”,
“message”: “File with such name does not exist.”,
“status”: 404
}
and I can't download it. I'm getting the same error with several other models as well.
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.