Information
This is the Krea 2 version of my Barbarian Warrior outfit from Conan Exiles Legacy. The Anima version can be found here.
Krea 2 model pages will be longer. I cannot write short and sweet for Krea 2. You may see the cover image be something slightly different. It's on purpose.
Disclaimer
If you generate imagery with my content, and you'd like to share it, please consider using the "Add Post" button at the bottom of the gallery instead of Post Images button at the top of the site, it gives the creator buzz and it costs nothing on your end. The article that was linked is a very important one I recommend reading.
I mainly create for fun, but I would appreciate it regardless.
Prompt Usage
For my Krea 2 models, this is divided into two sections. The first is the name of the outfit, one of the dataset images, and its trained prompt. The second is a random prompt from the gallery.. a little easier to work with.
I do not recommend people use the first prompt, at least as is.
Barbarian Warrior - Dataset Prompt (in-game preview)
b4rb4w4rr1, A woman with fair skin stands at a three-quarter angle, her body turned slightly so both front and side are visible, arms resting at her sides. Thin black eyeshadow lines her eyelids. She has short black hair. She wears a circlet, a necklace, a dark brown leather bra, a gold armband, wrapped forearm bands, a silver ornate belt, a brown wrap skirt, a side slit on the skirt, a hip pouch, leg wraps, brown boots. The dark brown leather has a matte finish, the gold armband shows a warm metallic sheen, and the silver belt has an ornate polished surface. The framing is a full-body shot that shows her entire figure from head to feet. The camera is positioned at a three-quarter eye-level angle. She is outdoors in a grassy field under open sky. The ground is covered in grass with no other structures nearby. Bright natural daylight falls evenly across the scene, picking out the metallic surfaces of the gold armband and silver belt while casting soft shadows beneath her arms and along the folds of the leather and fabric. The image is a realistic 3D render from a video game. It shows
detailed skin textures and material textures.Barbarian Warrior - Gallery Prompt
b4rb4w4rr1, A woman with fair skin. She stands facing the camera in an upper-body view, cropped at the waist, with a hand resting on her cheek, looking at the viewer. She has black eyes with black hair that is styled in a long braided ponytail. She is blushing, with a seductive smile. She wears a circlet, a necklace, a dark brown leather bra, a gold armband, and wrapped forearm bands. The frame is an upper-body view, cropped at the waist. She is outdoors in a floating city in the sky, stone platforms underfoot, buildings hanging in the air, and no grass. dark_art_style, retro anime hentai style. The image is a 2.5D retro anime illustration with soft modeled shading, rounded forms, and a slight depth between the character and the background. Linework is clean but not flat, colors are saturated, and the rendering has the soft airbrushed look of older hentai animation rather than hard cel shading. <lora:Flat_anime_Krea_@:0.8> <lora:barbarianwarrior-krea2-epoch-10:0.8>Styles and Checkpoints
I take off the last essay of my original prompt relating the outdoors grass and stuff, and replace it with a style wildcard that has a mix of LoRAs and non-LoRAs, and I run them on an x/y/z plot on three checkpoints.
The checkpoints are the Official Krea 2 Turbo, Lustify! v10 and Wulver v0.5. The int8 versions, preferably. My dataset is trained off vidya game images and I'm not a fan of realism, so your best chance is using the second one or something ideal for photorealism.
As for styles.. well.. Here is a brief explanation of what the styles wildcard will generate.
Anima Flat Style (it isn't actually flat and I'm not a fan)
Dark Fantasy Anime (my favorite)
ta86_inkriot (Borderlands style, it's solid)
Warhammer: Rogue Trader style (it's alright)
The purple retro anime style from the official turbo model, it's.. okay
A photorealism line, an attempt but it's not my focus
The trained dataset line, the 3d render look.
Training Parameters
All my outfits contain 50 images with varied poses, angles and frames.
I either train them on site at 10 epochs (2000 steps) with default settings, or locally at 512x512 resolution at 10 epochs (1000 steps). Typically, I will hand out the last and final epoch, because I use constant scheduler.
Description
Same as before, but the 10th epoch.