Instructions to use distilan/sb-image-loras with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Diffusers
How to use distilan/sb-image-loras with Diffusers:
pip install -U diffusers transformers accelerate
import torch from diffusers import DiffusionPipeline # switch to "mps" for apple devices pipe = DiffusionPipeline.from_pretrained("OpenMOSS-Team/MOSS-Image-SFT", dtype=torch.bfloat16, device_map="cuda") pipe.load_lora_weights("distilan/sb-image-loras") prompt = "stelleve9, evieblade6" image = pipe(prompt).images[0] - Notebooks
- Google Colab
- Kaggle
- Local Apps Settings
- Draw Things
- DiffusionBee
Stellar Blade EVE & Evie LoRAs
These are character LoRAs for EVE and Evie from Stellar Blade (+ Blood Rain DLC). Three image models: Krea 2, MOSS-Image and Anima.
Trigger words
| Character | Trigger word |
|---|---|
| EVE | stelleve9 |
| Evie | evieblade6 |
Just drop the trigger word into your prompt wherever you'd name the character, like stelleve9 sitting on the edge of a bed... or evieblade6, short hair, standing in a studio....
Want both girls in one shot? Load an EVE LoRA and an Evie LoRA together and use both triggers (check out the "Girls on vacation" prompt below).
If the hairstyles look off, try adding "black hair, long ponytail with bangs" for EVE and "dark brown, short hair" for Evie in your prompt.
Which file goes with which model
Each base model gets its own set of LoRAs, so grab the file that matches the model you're running.
| File | Character | Base model | Status |
|---|---|---|---|
stelleve9-krea2.safetensors |
EVE | Krea 2 | Tested on Krea 2 Turbo. Should work on Krea 2 Raw too. |
evieblade6-krea2.safetensors |
Evie | Krea 2 | Tested on Krea 2 Turbo. Should work on Krea 2 Raw too. |
evieblade6-moss.safetensors |
Evie | MOSS-Image | Works great with MOSS-Image-SFT, on its own or with the lodestones 8-step DMD LoRA. Should work with MOSS-Image-Base too. |
stelleve9-anima.safetensors |
EVE | Anima | Tested and known to work with Anima Base, Anima Aesthetic, and Anima Turbo. |
Quick version:
- Krea 2: EVE and Evie
- MOSS-Image: Evie only; EVE is already in the model's training data
- Anima: EVE only
Strength
- Run it at 1.0 if you want the game-accurate art style.
- Want something different, like anime, DC / Marvel comic books or watercolor? Try lower strengths like 0.8, 0.6 or even 0.4, and describe the style in your prompt.
This works the same way on all three models.
How to use it (ComfyUI)
Easy method: drag and drop one of the included images here in a ComfyUI window.
OR
- Download the
.safetensorsfile for your model and put it inComfyUI/models/loras/. - Load your base model workflow like you normally would (Krea 2 Turbo, MOSS-Image-SFT or Anima Base).
- Double click in any empty space on the workflow.
- Add a LoraLoaderModelOnly node between the model loader and the sampler, then pick the LoRA file and set the strength (see above).
- Using other LoRAs? Give each one its own LoraLoaderModelOnly node and chain it with these LoRAs.
- Write your prompt with the trigger word (
stelleve9orevieblade6) and hit generate.
Stick with your base model's usual settings.
Model-specific caveats
- Krea 2 (highly recommended): use it with FilterBypass V3 LoRA. It's a "decensoring" LoRA in theory, but it vastly improves Krea's prompt adherence, even with SFW images. Use str 0.8 - 1.0 depending on how source-accurate you want Evie or EVE to be.
- MOSS-image: I only included Evie LoRA for this model because EVE is already in its training data. Works great in tandem with the lodestones 8-step DMD LoRA. Use str 0.4 - 1.0 depending on how source-accurate you want Evie to be.
- Anima: Works best with str 0.4 - 0.8. Don't expect game-like visuals from this model because it's specifically trained to not generate photorealism. Stick to the model's strengths.
Sample prompts
EVE (stelleve9), Krea 2 Turbo
candid night flash photography, stelleve9 sitting upright, legs crossed over on the edge of a bed in a bedroom, smiling, winking, looking at the viewer, one hand raised in two victory fingers sign, black hair, ponytail with bangs, white crop top and torn blue denim shorts, white tennis shoes
The cheongsam sample in the header uses a twist on this same prompt:
night flash photography, stelleve9 sitting upright, legs crossed over on the edge of a bed in a bedroom, smiling, winking, looking at the viewer, one hand raised in two victory fingers sign, black hair, long ponytail with bangs blue and red cheongsam with vibrant colorful floral pattern (pink, yellow, blue flowers and green leaves), black strap heeled sandals,
Disney style:
Disney Ultradetailed Illustration, stelleve9 sitting cross legged in a bedroom, smiling, winking, looking at the viewer, one hand raised in separated victory fingers sign, black hair, long ponytail with bangs blue off-shoulder ruffled crop top with short sleeves and small buttons down the front, frilly ruffles on the neckline and hem, flowing white low waist skirt, visible belly button, with vibrant colorful floral pattern (pink, yellow, blue flowers and green leaves), black strap heeled sandals,
Chibi:
chibi anime style, stelleve9 standing straight in a desert, black hair, long ponytail with bangs, smiling, winking, looking at the viewer, victory fingers, green outfit
Evie (evieblade6), Krea 2 Turbo
Place and lighting:
photoshoot, flash photography, night time, building rooftop
Pose and framing:
evieblade6, short dark brown hair, light freckles, is standing on a building rooftop, viewed from a slightly low three-quarter front angle. Her torso is pressed against a full size wall behind her. Her right hand is up holding a white cigarette, white fumes, left hand hanging to the side, looking to the right
Outfit:
school uniform, white collared shirt, red striped bowtie, red plaid skirt, short sleeves, white loose socks, black school shoes
Anime illustration with a red halo (strength 1.0):
stylized anime illustration of a evieblade6 in profile, seated with her hands braced behind her. she wears a shiny black sleeveless dress. a bright red glowing ring forms a halo-like circle behind her head against a dark red background.
Fisheye selfie on a train platform (strength 0.8):
evieblade6, full body, dynamic selfie composition, (extreme high-angle view:1.3), (wide-angle lens:1.3), fisheye perspective, dramatic foreshortening, camera held above and close to her face, one arm extended toward the camera, other arm stretched sideways for balance, leaning forward playfully, looking directly at viewer, cheerful mischievous smile, short dark hair covering her left eye, light freckles, school uniform, white collared shirt, red striped bowtie, red plaid skirt, short sleeves, white loose socks, brown loafers, thick white colored intricate gauntlets gloves with studs and bolts, standing on a Korean railway station platform, train stopped along the left side, railway tracks on the right, platform roof, fluorescent ceiling lights, overhead electrical lines, metal beams, station signs blurred in the distance, yellow tactile paving, painted concrete floor, bright morning sunlight streaming across the platform, warm rim light, long soft shadows, subtle lens flare, shallow depth of field, cinematic depth, polished modern anime illustration, delicate painterly shading, clean linework, soft pastel colors, lively youthful atmosphere,
EVE + Evie together, Krea 2 Turbo
"Girls on vacation", with both character LoRAs loaded:
Watercolor Anime Illustration, near a swimming pool, evieblade6, short hair, standing, white monokini, smiling, looking at the viewer, hugging stelleve9 from behind. stelleve9 has dark hair with bangs and a ponytail, is wearing a green monokini, smiling, winking, sitting on a recliner and sipping coke with a straw, one hand doing the victory V gesture, looking at the viewer
Evie (evieblade6), MOSS-Image
Fashion photoshoot (you'll also find this one as the repo sample images/moss-image_00194_.png):
evieblade6, short hair, with fair, realistic, natural skin, standing in a confident fashion pose in a minimalist studio with a smooth dark gray to light gray gradient background. She is wearing oversized white-framed rectangular sunglasses. She is wearing a tight camouflage-patterned sheer sleeveless tank top with a square neckline that accentuates her figure, layered with multiple delicate silver necklaces including star and cross pendants. Her high-waisted wide-leg metallic silver pants have a glossy, reflective chrome finish, with visible stitching, a silver chain accessory hanging from the waistband, and her hand casually in one pocket. She wears black open-toe sandals with ruched fabric straps and clear chunky heels. Pose: body slightly angled, one hand resting on her thigh, the other near her hip, head tilted right with a cool, aloof expression. Soft studio lighting, high fashion editorial style, sharp details, cinematic lighting, 8k resolution.
Comic book style:
marvel DC comics aesthetic, flat 2d, evieblade6 hanging off a street light with one hand, new york city in background, short dark hair, smiling, winking, looking at the viewer, one hand in victory fingers, spider man outfit
Chibi anime:
chibi anime aesthetic, evieblade6 standing straight in a desert, short black hair, smiling, winking, looking at the viewer, one hand with victory fingers, white outfit
Portraits:
evieblade6, close-up head and shoulders portrait, short dark hair, fair realistic natural skin with visible pores and fine detail, looking directly at the camera with a calm neutral expression, plain black crew-neck top, seamless soft gray studio background. Soft even beauty-dish lighting, shallow depth of field, 85mm lens, sharp focus on the eyes, photorealistic, high detail, 8k resolution.
evieblade6, three-quarter view portrait from the chest up, short dark hair slightly tousled, fair realistic natural skin, head turned slightly toward the light with a soft hint of a smile, wearing a simple cream knit sweater, sitting by a large window in a quiet apartment. Warm natural window light from the side, gentle shadows on the face, blurred background, candid editorial photography, 50mm lens, photorealistic, sharp details.
evieblade6, side profile close-up portrait, short dark hair tucked behind the ear, fair realistic natural skin, serious focused expression, looking off into the distance, small silver stud earring, dark background. Low-key cinematic lighting with a single rim light outlining the jaw and nose, high contrast, moody, film grain, photorealistic, extremely detailed face, 8k resolution.
That last one made this portrait (Evie at 0.9 plus the 8-step DMD LoRA at 0.7):
Tip: MOSS-Image loves long, detailed prompts. Phrases like "realistic natural skin" and "photorealistic, extremely detailed face" push it toward realism.
Download
Grab the files from the Files and versions tab.
- Downloads last month
- 79
Model tree for distilan/sb-image-loras
Base model
OpenMOSS-Team/MOSS-Image-Base