This is a textual inversion embedding for Aerith Gainsborough from Final Fantasy VII (essentially both original and remake together).
As many of her appearances show the pink dress in a very light color, it's best to prompt this with green eyes, pink dress, corneo_aerith as it sometimes (though rare) doesn't get the dress or eye color right by itself.
v2 example images produced with my 7th Heaven mix https://civarchive.com/models/4669/corneos-7th-heaven-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)!
I now have a running poll for the next character embed to work on at https://strawpoll.com/master-corneo
You can comment with any requests to add but it needs to be someone with enough fan art available to train on, and also a character I'd want to do!
Description
FAQ
Comments (8)
I've got to be doing something wrong here because I can't get results like yours even with the same prompt, model, and settings using Automatic1111. I put embedding file in /embeddings folder, used orangecocoa5050mix (which gives really faded results, am I missing a VAE or something?) and copied all settings from two of your example images. Apologies if I missed something obvious and thank you in advance for any advice.
Faded results is pretty much always not using a VAE yes, I use the Anything v3 VAE with everything. I've realised now there's a hypernetwork on these, though I don't know exactly what it is or what it does, so I'll do the same you have and try recreating these based on what civitai is saying their settings are and see what happens
Thank you very much. Would I copy and rename the Anything V3 VAE to match the model's name that I want to use it with then? I'm still getting the hang of using anything other than basic checkpoint models, so it's entirely possible I'm just missing something. Appreciate your time.
Just tested it now and I'm still getting the same (at least tried out the first image listed). One other thing it probably is is that civitai doesn't list the image size, this one's 512x768. Better is opening the image in a new tab, dropping it in PNG Info in Automatic1111's web UI and then txt2img which should be working
Ah and with the VAE, it's better to go into the settings, and assuming you're up to date (recent git pull - though be warned this does seem to break PNG info for the moment), go to User interface > Quicksettings list and add `sd_vae` into a comma separated list there. That gives you a VAE selection box at the top of the screen for easy flipping between them or turning it off. My complete Quicksettings list is `sd_model_checkpoint, sd_hypernetwork, sd_vae, CLIP_stop_at_last_layers `
Thank you, I am actually getting better results now because of your Quicksetting list. I needed to change clip skip to 2 like in your first example, and although the result isn't identical to your first example, it's really close. Could you explain why clip skip has such an effect when you have a moment? I'll do some more testing and leave a 5-star review with the examples soon. :)
I don't see corneo_arm_grab_missionary embedding on here. Where can it be found?
It's one of several embeds that don't quite work well enough so I've not released it, but now I've figured out a problem in my training data I'm hoping retaining it will sort it
Details
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.