BLOCKv0.5

BLOCKv0.5 is an image-to-image model for converting a 3D Minecraft character preview into a 2D skin texture file.

It is based on black-forest-labs/FLUX.2-klein-base-9B with merged LoRA weights and was trained with [cond, target] token order.
For best results, use the included custom pipeline: Flux2KleinPipelineTrainOrder.

What This Model Does

• Input: a character preview image (RGB, square recommended)
• Optional control: text prompt for style/details
• Output: a generated skin texture image (PNG)

Quick Start

import torch
from PIL import Image
from pipeline_flux2_klein_train_order import Flux2KleinPipelineTrainOrder

model_id = "your-username/BLOCKv0.5"  # or local path

pipe = Flux2KleinPipelineTrainOrder.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
)
pipe.to("cuda")

preview = Image.open("examples/ex1_preview.png").convert("RGB").resize((1024, 1024))

result = pipe(
    prompt="clean minecraft skin texture, readable details, game-ready layout",
    image=preview,
    num_inference_steps=30,
    guidance_scale=4.0,
).images[0]

result.save("generated_skin.png")

Example Results

Preview	Generated Skin

Notes

• This checkpoint is designed for img2img skin generation, not general text-to-image generation.
• The included custom pipeline matches the training token order and is recommended for inference.
• The model is large; use a high-memory GPU for practical inference speed.

Citation

If you use BLOCKv0.5 or results derived from this model, please cite:

@article{guo2026block,
  title={BLOCK: An Open-Source Bi-Stage MLLM Character-to-Skin Pipeline for Minecraft},
  author={Guo, Hengquan},
  journal={arXiv preprint arXiv:2603.03964},
  year={2026},
  url={http://arxiv.org/abs/2603.03964}
}

Dependencies

• diffusers
• transformers
• torch
• safetensors