Credits and licenses

MOSS-Image

Blokjutsu is an independent app for MOSS-Image, the open-weight text-to-image model and pipeline by the OpenMOSS Team and MOSI.AI. All of the picture-making is theirs. Blokjutsu is not affiliated with or endorsed by them.

If you use pictures or results from Blokjutsu in research, please cite MOSS-Image:

@misc{mossimage2026,
  title        = {MOSS-Image},
  author       = {{OpenMOSS Team}},
  year         = {2026},
  howpublished = {\url{https://github.com/OpenMOSS/MOSS-Image}}
}
@misc{mossimage2026report,
  title         = {MOSS-Image Technical Report},
  author        = {{MOSI.AI}},
  year          = {2026},
  eprint        = {2610.23374},
  archivePrefix = {arXiv},
  url           = {https://arxiv.org/abs/2610.23374}
}

Models

MOSS-Image’s model is published on Hugging Face as MOSS-Image-SFT (and the untuned MOSS-Image-Base). Its text encoder and image decoder, from Qwen and FLUX.2, are included in that download.

model by license
MOSS-Image-SFT, MOSS-Image-Base OpenMOSS Team Apache-2.0
Qwen3.5-9B: MOSS-Image’s text encoder; its output layer is downloaded separately for describing Qwen Team, Alibaba Cloud Apache-2.0
FLUX.2 VAE: MOSS-Image’s image decoder Black Forest Labs Apache-2.0

Libraries

PyTorch and torchvision (BSD-3-Clause), Diffusers, Transformers, Accelerate, huggingface_hub, and safetensors (Apache-2.0, Hugging Face), PySide6 / Qt (LGPL-3.0), Pillow (MIT-CMU), NumPy and httpx (BSD-3-Clause), PyObjC (MIT), and uv (Apache-2.0 / MIT, Astral) for installing. This guide is built with Jekyll (MIT) and the Just the Docs theme (MIT).

Open source

Blokjutsu, MOSS-Image’s code, and MOSS-Image’s model weights (including the Qwen3.5-9B text encoder and FLUX.2 image decoder inside them) are all released under the Apache License 2.0. Anyone may use, study, change, and share them, for any purpose, commercial or not, provided the license and copyright notices are kept and changes are noted. The license also grants the patent rights needed to use them.

Pictures in this guide

The ice cream pictures and the examples were made with Blokjutsu. The baseball card used as a starting picture in Examples is [Cy Young, Cleveland Naps, baseball card portrait], card set White Borders (T206), issued by the American Tobacco Company, 1909–1911, from the Library of Congress, Prints and Photographs Division (Benjamin K. Edwards Collection), with no known restrictions on publication (loc.gov/pictures/item/2008676577).

The photograph in Describing images is “Sunday cyclists in East Potomac Park, Washington, D.C.” by Marjory Collins, June–July 1942, from the Library of Congress’s Farm Security Administration/Office of War Information collection, which is in the public domain (loc.gov/item/2017825632).

Blokjutsu

Written by Bob Carpenter together with Claude, Anthropic’s AI model. Copyright 2026 Bob Carpenter, Apache License 2.0. Two files come from MOSS-Image’s own code (also Apache-2.0): blokjutsu/moss_image.py, copied unchanged, and blokjutsu/moss_forward.py, adapted from it. To cite Blokjutsu itself, use the “Cite this repository” button on GitHub.

“MOSS-Image” is the OpenMOSS Team’s name and is used here only to say what Blokjutsu runs.

Etymology

Blokjutsu (BLOK-joot-soo, /ˈblɒk.dʒuː.tsuː/), noun. The art of the block.

From Danish blok, “block,” and Japanese jutsu (術, じゅつ), “art” or “technique.” The name comes from how MOSS-Image works: it treats an image as blocks of pixels, each 16 × 16 square one unit for the model, and transforms them step by step from noise into the finished image.

Design inspiration

The block designs in the icon and the app draw on

  • the Kohs block design test, introduced by Samuel C. Kohs in 1920: colored cubes, some faces split diagonally, arranged to match a pattern
  • Froebel’s gifts, the wooden building blocks of Friedrich Fröbel’s first kindergartens
  • the primary colors and simple geometry of the Bauhaus