Trained on images based on her Resident Evil 1 remake design.
Illustrious usage and notes:
Positive prompt: jill valentine, beret, 1girl, blue eyes, short sleeves, breasts, pants, boots
Additional tags: watch, fingerless gloves
For full prompt examples, look at metadata for the image uploads.
PonyXL usage and notes:
Positive prompt: 1girl, RE1Jill, blue eyes, brown bob cut, beret, breasts, fingerless gloves
Optional tags to emphasize parts of the design if they don't show up: shoulder pads, sleeves, pants
Add watch to positive or negative prompt if that's something you want added or avoided.
For full prompt examples, look at metadata for the image uploads.
For an alternate PonyXL take on Jill from RE1, Lamancha50002310 has made a model with Jill (RE3 and RE1 designs) here: https://civarchive.com/models/435001/jill-valentine-l-ponyxl-2-outfits
Description
FAQ
Comments (10)
Any chance for u add RE3 outfit in this jill please?
I have an RE3 Jill I'll be uploading. The dataset is entirely different, though, so the model won't have the exact same style as this one.
Uploaded RE3 Jill model: https://civitai.com/models/1415070/jill-valentine-resident-evil-3-illustrious
@WizardMcMerlin I think she looks great in this model, less so in the previews for the re3 one unfortunately where she looks a bit younger and more fresh faced/baby faced. Just my 2 cents
@doveship Yeah, that's the result of the very different dataset I used. I'm not super happy how the face turned out either, but I did it as part of a commission and the client was happy with it so that's good enough for me.
alcina dimitrescu please!
re5 outfit pls pls pls!!!
How many pictures did you used for training, and what kind of them? I mean did you used zoom-in, specificated folders for beret, face or something like this, or only used just full body pictures
There were 83 images in the dataset in a mix of styles. All images were based on her RE1 remake design. I didn't do much to the images. I'm not even sure if I cropped any of them.
To be honest, you can get away with a lot. When I did my DMC anime LoRA, I simply took screengrabs from the anime and did some basic captioning. I didn't crop a single one, and that was despite the character not always being in the center or the primary subject in an image. And every image was the same widescreen aspect ratio. But the LoRA still turned out good.
One thing I haven't quite figured how to master is faces with Illustrious LoRAs. If I create a LoRA that's meant to have a more realistic style, or 3D-focused style, then there's a chance faces will end up looking off, even if I train for longer and includes many images that are zoomed-in on the face.
@WizardMcMerlin Oh, I see you have 1 repeat per image. So... Grats, this is the best Jill model on this site!
Details
Files
Available On (1 platform)
Same model published on other platforms. May have additional downloads or version variants.