Moss-Image Opens Its Weights with an Apache 2.0 License and Working Japanese Text
A permissively licensed model that renders Japanese text cleanly removes two common barriers for commercial work, licensing restrictions and garbled lettering, so the practical question shifts from whether open models can do this to how quickly small studios adopt one.
Reporting from 1 source: ASCII.jp.
Moss-Image, a 5B-parameter image model, was released as open weights on October 16. It generates images from photographic to anime style and renders text, including Japanese. The model uses the Apache 2.0 license, which permits commercial use, modification, redistribution, and embedding in products. It was developed by the Open MOSS Team, a research organization affiliated with universities in Shanghai, with development funding from the Shanghai city government through related companies. LoRA support is already being put in place.
The release arrived with Apache 2.0 attached, which lets a shop or a small studio modify the weights and ship them inside a product without negotiating a separate license. The model runs at 5B parameters, sizeable enough for varied output but not the scale that demands a cluster.
For a test, the team analyzed an image of a character called Asuko Asu and had GPT-6 Astra draft prompts for a simple tourism poster for a fictional town. They made a photographic version, swapped the positions of the people, then produced an anime-style version of the autumn travel advertisement. Japanese text came out solidly in both.
Written with AI by Yomimono from the 1 cited source below, including Japanese-language reporting where cited. See our editorial policy.
Add Yomimono as a preferred source on Google