Mineskin Diffusion v1.0

A quick and high-quality model for creating your owns Minecraft Skins

This a new and rebuild version of old Minecraft Skin Diffusion models. Model have about 18 millions params, that make it a very fast and don't need a lot of memory. Model train in 30000 images of mans and girls skins.

Result of model

1. Example 1
2. Example 2

Usage

You can run this model in Google Colab notebook. This notebook provide all instrumens for comfortable using this model and provide 3DSkinView for looking skin in 360 view. If your want use this model on your local machine use this steps:

1. You should have this library numpy, transformers, diffusers, pillow, torch;
2. Use this pipeline for faster inference:

   device = ("mps" if torch.backends.mps.is_available() else "cuda" if torch.cuda.is_available() else "cpu")
   
   class MSPipeline(DiffusionPipeline):
     def __init__(self, unet, scheduler):
        super().__init__()
        self.register_modules(unet=unet, scheduler=scheduler)
   
      @torch.no_grad()
      def __call__(self, batch_size = 1, num_inference_steps = 1000):
        x = torch.randn(batch_size, self.unet.config.in_channels, self.unet.config.sample_size, self.unet.config.sample_size).to(device)
   
        self.scheduler.set_timesteps(num_inference_steps)
   
        for t in self.progress_bar(self.scheduler.timesteps):
          model_output = self.unet(x, t).sample
          x = self.scheduler.step(model_output, t, x).prev_sample
   
        x = x.cpu().permute(0, 2, 3, 1).clip(0, 1).numpy() #* 255
        x = self.numpy_to_pil(x)
   
        return x
   

3. Download model using: pipe = MSPipeline.from_pretrained("WiNE-iNEFF/Mineskin-Diffusion-v1.0", use_safetensors=True).to(device);
4. Model can work with: DDIMScheduler, DDPMScheduler, DEISMultistepScheduler, LCMScheduler, PNDMScheduler, UniPCMultistepScheduler:

   #Example of using DDIMScheduler
   pipe.scheduler = DDIMScheduler.from_pretrained("WiNE-iNEFF/Mineskin-Diffusion-v1.0", subfolder="scheduler")
   

5. Run inference with using: images = pipe(batch_size = 1, num_inference_steps = 30);