README.md

metadata

license: wtfpl
base_model: black-forest-labs/FLUX.1-dev
pipeline_tag: text-to-image
tags:
  - text-to-image
  - lora
  - diffusers
  - template:diffusion-lora
  - flux
  - flowmatch
  - stationthis
widget:
  - text: >-
      nanase, This is a colored pencil drawing of a young girl with light brown,
      shoulder-length hair, smiling widely with her eyes closed. She has fair
      skin and is depicted in a school uniform, wearing a light blue shirt with
      a blue and white striped tie. Her left hand is raised to her head,
      adjusting her hair. The background features soft, green and yellow
      brushstrokes, adding a sense of warmth and happiness to the image. The
      girl's expression is joyful and carefree. The artist's signature, "YUMI
      2020," is in the bottom left corner. The overall style is cute and
      anime-inspired.
    output:
      url: samples/sample_000.jpg
instance_prompt: nanase
training_steps: 4000
network_type: lora
library_name: ai-toolkit

nanase

This LoRA model generates colored pencil style illustrations featuring young girls with light brown, shoulder-length hair and fair skin, often depicted in school uniforms. The visuals capture joyful expressions, with characters typically smiling widely with their eyes closed, adding a playful and charming aesthetic.

Trigger word: nanase

Sample Outputs

nanase, This is a colored pencil drawing of a young girl wit...	nanase, This is a digital watercolor-style anime drawing of ...
nanase, This is a traditional watercolor and pencil drawing ...	nanase, This is a watercolor and ink drawing of an anime-sty...

Usage

ComfyUI

1. Download the .safetensors file from the Files tab
2. Place in ComfyUI/models/loras/
3. Use the Load LoRA node with strength 0.8-1.0
4. Include nanase in your prompt

Diffusers

import torch
from diffusers import FluxPipeline

pipe = FluxPipeline.from_pretrained(
    "black-forest-labs/FLUX.1-dev",
    torch_dtype=torch.bfloat16
)
pipe.load_lora_weights("ms2stationthis/nanase")
pipe.to("cuda")

image = pipe(
    prompt="nanase portrait, soft lighting, detailed",
    guidance_scale=3.5-4.0,
    num_inference_steps=20-30,
    generator=torch.Generator("cuda").manual_seed(42)
).images[0]

image.save("output.png")

Recommended Settings

Parameter	Value
LoRA Strength	0.8-1.0
Guidance Scale	3.5-4.0
Inference Steps	20-30
Resolution	1024x1024

Example Prompts

• nanase shoulder-length hair
• nanase adjusting her hair. the background features soft
• nanase green and yellow brushstrokes
• nanase "yumi 2020

Training Details

• Base model: black-forest-labs/FLUX.1-dev
• Training steps: 4000
• Trigger word: nanase
• Network: LoRA rank 32, alpha 32
• Optimizer: adamw8bit, lr 1e-4
• Precision: float16
• Resolution: 512, 768, 1024 (multi-res bucketed)

About

Trained using StationThis — an AI creative platform powered by $MS2. Train your own LoRAs via @stationthisbot on Telegram.

---

_{Generated by StationThis Training Pipeline}