One idea, four ways to write it
There is no single right way to write an image prompt. FLUX reads sentences. Stable Diffusion 1.5 likes short comma-separated phrases and an avoid list. Anime models think in tags. MOSS-Image wants a detailed caption with the exact words in quotes. Writing for the wrong model wastes renders.
How to fix it
- Photo (FLUX): write a photo brief. Subject and action, setting, then camera (close-up, 85mm, shallow depth of field), light (soft window light from the left) and mood. Plain sentences. No "8k, masterpiece".
- Illustration (DreamShaper): short phrases plus style. "a fox reading under a mushroom, rain, fireflies" plus a style such as watercolour or storybook, and a list of things to avoid.
- Anime (MOSS tags): quality, character, then details. "best quality, highres, 1girl, silver hair, school uniform, cherry blossoms, looking back". Count tags like 1girl matter.
- Text & Posters (MOSS): medium and subject first. "Vintage travel poster of ...", then layout, light, colour, and every word to print in quotes.
- Or let AI write it. Every tab has an AI helper that rewrites your idea in the style that tab's model understands best. Undo is one click.
Try it
The Photo tab is set up for exactly this, with examples to start from and an AI helper that writes the prompt for you.