AliceKJ/BLOCKv0.5

BLOCKv0.5

This refreshed BLOCKv0.5 release continues the original FLUX.2-klein-base-9B line and keeps the original train-order inference path.

Compared with the earlier BLOCKv0.5 release, this update merges a newer 9B checkpoint with further iteration and fixes on the 9B model line. If you want the newer 4B release that works directly with the official Flux2KleinPipeline, see BLOCKv0.6.

What This Model Does

• —Input: a character preview image with front/back views
• —Optional control: text prompt for style/details
• —Output: a generated Minecraft skin UV atlas

Important Note About Inference

BLOCKv0.5 still follows the original train-order inference contract. Use Flux2KleinPipelineTrainOrder for inference with this checkpoint.

Quick Start

import torch
from diff_mc.pipelines import Flux2KleinPipelineTrainOrder
from PIL import Image

model_id = "AliceKJ/BLOCKv0.5"

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

preview = Image.open("your_cond.png").convert("RGB").resize((512, 512))
result = pipe(
    prompt="Reference-based image-to-image. Input is a front+back 3D Minecraft character reference image. Output a Minecraft skin UV texture atlas (64x64 pixel art layout), matching the character in the reference. High-quality, anime-inspired. Clean pixel art, sharp edges, no blur, no anti-aliasing. Keep consistent UV placement and mapping; match the same character design from the reference. Model type: slim.",
    image=preview,
    num_inference_steps=30,
    guidance_scale=4.0,
).images[0]

result.save("generated_skin.png")

Example Results

The example images below are preserved from the earlier BLOCKv0.5 release for continuity.

Notes

• —BLOCKv0.5 remains on the FLUX.2-klein-base-9B branch.
• —This updated release reflects further iteration and fixes on the 9B model line.
• —BLOCKv0.5 still uses Flux2KleinPipelineTrainOrder.
• —For the 4B official-pipeline release, see BLOCKv0.6.

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}
}