Sabikaze for MOSS-Image: a quiet anime illustration style with a desaturated grey-blue / grey-teal palette, one small warm accent, heavy ink lines, flat value blocks and dry-brush texture on clothes and backgrounds. It is the MOSS-Image version of the Anima and Krea 2 LoRAs "Sabikaze", trained on the same 57 self-generated images.
How to prompt: start with "@z3-sabikaze." and name the medium in a sentence, for example "@z3-sabikaze. Anime-style digital illustration of ...". Without a sentence that says how the picture is drawn, MOSS-Image renders a photograph. Describe the scene in plain sentences; a figure placed in a real environment (streets, rooms, stalls, harbors) works best, with one warm accent named on something in the scene.
Strength: tested at 1.0. Once the LoRA is loaded the style also applies without the trigger.
Tested with ComfyUI 0.42.1, MOSS-Image SFT (int8 ConvRot DiT), the Qwen3.5 9B text encoder, 25 steps, CFG 2.5, er_sde / simple, shift 5, 832x1248, loaded with LoraLoaderModelOnly.
Training: MOSS-Image SFT with ai-toolkit, 57 images, rank 32, 2000 steps, captions with the trigger first and the medium named.
Known limit: scenes with shop fronts or walls sometimes get a sign with made-up text, even with "No text in the image." in the prompt; try another seed.