File size: 2,917 Bytes
ea99535 caac083 ea99535 caac083 ea99535 caac083 ea99535 caac083 ea99535 9c1b82c caac083 9c1b82c caac083 9c1b82c caac083 9c1b82c caac083 9c1b82c caac083 9c1b82c caac083 9c1b82c caac083 9c1b82c 6d06c33 9c1b82c f9a0ac2 9c1b82c caac083 9c1b82c fc7246f | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 | ---
license: apache-2.0
library_name: diffusers
pipeline_tag: image-to-image
base_model: black-forest-labs/FLUX.2-klein-base-9B
tags:
- flux2
- img2img
- minecraft
- skin-generation
- diffusers
- safetensors
---
# 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](https://huggingface.co/AliceKJ/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
```python
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.
| Preview | Generated Skin |
| --- | --- |
|  |  |
|  |  |
|  |  |
## 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](https://huggingface.co/AliceKJ/BLOCKv0.6).
## Citation
If you use BLOCKv0.5 or results derived from this model, please cite:
```bibtex
@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}
}
```
|