# mossimage-prompts # MOSS-Image Prompt Collection ## Gallery and reproducible examples - [MossImage gallery](https://mossimage.run/gallery/) — published images with prompts and saved settings. - [10 MOSS-Image prompt examples and seeds](https://mossimage.run/blog/best-moss-image-prompts-examples-seeds/) — bilingual menus, posters, labels, and JSON layout examples, with model labels and saved seeds. MossImage is an independent hosted studio, not affiliated with OpenMOSS. Generated lettering and layout should be checked before publishing. Tested and verified prompts for [MOSS-Image](https://arxiv.org/abs/2610.23374), the open-source 5B diffusion model. Every prompt in this repo has been run through MOSS-Image-SFT and confirmed to work. 👉 **[mossimage.run](https://mossimage.run)** — Try these prompts online, free, no setup. --- ## Text-to-Image ``` A serene mountain lake at sunrise, pine trees silhouetted against orange sky, photorealistic, 8K ``` ``` Minimalist Japanese bedroom, tatami mats, sliding shoji doors, soft natural light, architectural photography ``` ``` Close-up of a cat with heterochromia, one blue eye one green eye, soft fur texture, shallow depth of field, studio lighting ``` ``` A steampunk airship floating above a Victorian city at dusk, warm copper tones, detailed illustration ``` --- ## JSON Layout Control MOSS-Image-SFT lets you control object placement with JSON instead of prompt engineering. **Two objects:** ```json { "canvas": {"width": 1024, "height": 1024}, "objects": [ {"label": "red ball", "position": [0.2, 0.5], "size": [0.3, 0.3]}, {"label": "blue cube", "position": [0.8, 0.5], "size": [0.3, 0.3]} ], "style": "white background, studio lighting" } ``` **Coffee and book:** ```json { "canvas": {"width": 1024, "height": 1024}, "objects": [ { "label": "a cup of coffee on a wooden table", "position": [0.3, 0.5], "size": [0.4, 0.35], "detail": "ceramic white mug, steam rising, dark roast coffee" }, { "label": "an open book next to the cup", "position": [0.55, 0.45], "size": [0.3, 0.35], "detail": "old leather-bound book, yellowed pages" } ], "style": "warm morning light, cozy cafe atmosphere, shallow depth of field" } ``` **Three objects — tabletop still life:** ```json { "canvas": {"width": 1024, "height": 1024}, "objects": [ {"label": "a glass vase with a single sunflower", "position": [0.25, 0.4], "size": [0.2, 0.45]}, {"label": "a vintage brass pocket watch", "position": [0.55, 0.55], "size": [0.25, 0.25]}, {"label": "a stack of three old hardcover books", "position": [0.8, 0.45], "size": [0.3, 0.35]} ], "style": "warm afternoon light from a window, dust particles in the air, painterly realism" } ``` --- ## Chinese Text Rendering ``` 一张中文海报,标题"春日特卖"位于顶部,副标题"全场五折起",背景是樱花盛开,清新风格 ``` ``` A poster with Chinese text "你好世界" in bold calligraphy at the center, ink wash painting style, white background ``` ``` 一家奶茶店的店面招牌,写着"珍珠奶茶"四个大字,霓虹灯风格,夜晚街景 ``` --- ## English Text Rendering ``` A neon sign that says "OPEN" in pink and blue glow, dark brick wall background, cyberpunk aesthetic, rainy street reflection ``` ``` Product mockup of a white t-shirt with the text "LESS IS MORE" printed in Helvetica, front view, clean studio lighting ``` ``` Chalkboard menu with "TODAY'S SPECIAL" written in elegant chalk lettering, rustic coffee shop background, warm tones ``` --- ## Prompt Tips - **Layout accuracy:** JSON layout is ~80% accurate on SFT. Natural language alone is ~40%. Use JSON when position matters. - **Text rendering:** Tell the model what kind of text it is — "neon sign", "printed on paper", "chalkboard lettering". Context helps it render correctly. - **Prompt length:** Shorter prompts often work better. The built-in enhancer fills in details. Start with 10–30 words. - **Chinese + English mixed:** Quote the non-dominant language. The model handles mixed-language well. - **Fixed seed:** For reproducible results, use a fixed seed and square aspect ratio. Change one variable at a time. - **Model choice:** SFT for text rendering and layout. Base for general generation. --- ## Related - [mossimage-workflow](https://github.com/kaixin20190930/mossimage-workflow) — ComfyUI workflow JSON - [mossimage.run/comfyui/](https://mossimage.run/comfyui/) — ComfyUI setup guide - [MOSS-Image paper (arXiv)](https://arxiv.org/abs/2610.23374) - [OpenMOSS on GitHub](https://github.com/OpenMOSS) ## Contribute Submit a PR with your prompt. Include: category, model version (Base/SFT), a note on why it works, and an example image if possible. ## License CC0 — public domain. Free to use, no attribution required.