Phew... My first "serious" model...

Welcome to HARRY07979/SoraAnima-v1!

Demo space: https://huggingface.co/spaces/HARRY07979/Sora-Anima-v1

_{masterpiece, best quality, ultra-detailed, 1girl, long silver hair, blue eyes, blush, gentle smile, head tilt, white shirt, red ribbon tie, navy pleated skirt, sitting on chair, classroom, large window, golden hour, sunset sunlight, distant cityscape, warm soft lighting, bokeh}

Model Details:

• Developed by: HARRY07979(Me!)
• Model type: (Anime) Diffusion (or txt2img) model
• Base model: stable-diffusion-v1-5/stable-diffusion-v1-5 (Original base: runwayml/stable-diffusion-v1-5)

Quck Start:

from diffusers import DiffusionPipeline
import torch

# 1. Load the model onto the GPU (using float16 for standard SD 1.5 performance)
pipe = DiffusionPipeline.from_pretrained(
    "HARRY07979/SoraAnima-v1",
    torch_dtype=torch.float16,
    safety_checker=None,
).to("cuda")

# 2. Set Prompt & Negative Prompt
prompt = "masterpiece, best quality, 1girl, solo, cute smile, silver hair, school uniform, sitting by the window, sunset light, outdoors"

negative_prompt = "lowres, bad anatomy, bad hands, text, error, missing finger, extra digits, fewer digits, cropped, worst quality, low quality, low score, bad score, average score, signature, watermark, username, blurry"

# 3. Generate image using the recommended sweet-spot settings
image = pipe(
    prompt=prompt,
    negative_prompt=negative_prompt,
    num_inference_steps=30,  # Optimal steps for DPM++ Karras
    guidance_scale=6.5,  # Sweet-spot CFG (avoids harsh contrast and edge noise)
    width=512,
    height=512,
).images[0]

# 4. Save and display the image
image.save("soraanima_result.png")
image.show()  # Or use display(image) in Jupyter / Google Colab

Recommended Settings:

• Sampler: DPM++ 2M Karras (already)
• Sampling Steps: 25-30 (Recommended: 30)
• CFG Scale: 5.5-7.5 (Recommended: 6.5/7.5)
• Resolution: 512x512
• Clip Skip: 2

How-to Prompt

• Prompt:
  Use this following order to generate the best images:

[Quality/Style] + [Character Count] + [Character & Outfit] + [Action/Expression] + [Background/Camera Angle] + [Lighting]

Example:

masterpiece, best quality, 1girl, solo, cute smile, silver hair, school uniform, sitting by the window, sunset light, outdoors

• Negative Prompt (Recommended):

lowres, bad anatomy, bad hands, text, error, missing finger, extra digits, fewer digits, cropped, worst quality, low quality, low score, bad score, average score, signature, watermark, username, blurry