nikkijiang/sd-pokemon-generator

Model Description

This is a fine-tuned Stable Diffusion v1.5 model specialized for generating Pokémon-style character illustrations.
The model was trained using a custom dataset curated from public Pokémon image datasets and captions.

Model Details

• Base Model: Stable diffusion model for text-to-image generation[runwayml/stable-diffusion-v1-5]
• Fine-tuned on: Pokémon image-caption datasets including:
  • lambdalabs/pokemon-blip-captions
  • 42arch/pokemon-dataset-zh
• Model Size: This model consists of multiple files with a total size of approximately 4.3 GB

Usage

from diffusers import StableDiffusionPipeline
import torch

if torch.cuda.is_available():
    device = "cuda"
else:
    device = "cpu"
    
model_id = "nikkijiang/sd-pokemon-generator"

pipe = StableDiffusionPipeline.from_pretrained(model_id).to(device)

prompt = "a cute flying cat, shining eyes"
image = pipe(
        prompt='an official style pokemon :'+prompt,
        height=256,
        width=256,
        num_inference_steps=100,
        guidance_scale=7.5
    ).images[0]
image.show()