How to write a good image prompt for Midjourney, Stable Diffusion and DALL-E
Every AI image generator speaks a different language. Midjourney likes short keywords separated by commas. Stable Diffusion likes keywords too, but some with a number next to them that says *"this matters more"*. DALL-E prefers full sentences, like talking to a person. Flux sits somewhere in between.
You click what you want to see: subject, style (photorealistic, anime, watercolor and so on), mood, lighting, framing, color palette, image proportions. You can also list what to skip (*"no watermarks"*, *"no blurry background"*).
The tool builds a ready prompt in the format of your chosen generator. The same idea comes out in four versions. Paste it where you want, compare which one looks best. No need to remember how each service works.
How to use it
- What's in the picture: write briefly who or what. Example: *"a young witch on a broomstick"*. You can add what they're doing: *"casting a spell over a glowing cauldron"*.
- Pick a style from the list: photorealistic, cinematic, anime, oil painting, pixel art, cyberpunk, fantasy, minimalist, watercolor, 3D render. One click = a matching set of keywords.
- Tune the vibe: mood (dark, joyful), lighting (sunset, neon, candles), framing (low angle, face close-up), camera (85mm portrait lens, macro), palette (pastel, warm, black and white).
- Detail level: simple (fast, for sketching ideas), balanced (default), heavily detailed (slower, more detail, but sometimes overdoes it and messes up faces).
- Image proportions (e.g. 16:9, 1:1): 1:1 for Instagram, 9:16 for Reels and TikTok, 16:9 for desktop wallpapers or YouTube thumbnails, 4:5 for vertical posts.
- List of things to skip: click the defaults (blurry, watermark, ugly hands) or add your own. The tool puts them in the right place, every generator handles this differently.
- Pick where you'll paste it: Midjourney, Stable Diffusion, DALL-E or Flux. The format adjusts itself. You can compare all four versions side by side.
When this is useful
Six typical situations where this builder gives you a concrete edge:
- Testing the same idea in several tools. You have a concept (*"warrior in a forest at sunset"*). One click gives you four prompt versions (Midjourney, Stable Diffusion, DALL-E, Flux). Paste each one and pick the prettiest result.
- Making graphics for social media. You need an Instagram post (square 1:1), Reels (vertical 9:16), a blog header (horizontal 16:9). Click the format, the tool adds the rest. You don't need to remember that Midjourney uses --ar 16:9.
- Illustrations for a blog or e-book. You want every picture in the book to look similar. Pick one style (e.g. *"watercolor"*) and only change the subject of each. Everything looks like it came from the same hand.
- Playing with styles. You don't know what *"cyberpunk + photorealistic"* or *"fantasy + watercolor"* looks like. Click, generate, see. Learning by playing.
- Mockups for a client. The client wants three style options to choose from. You build three prompts, generate, present. Faster than drawing by hand.
- Learning how each generator works. You see how Midjourney differs from Stable Diffusion through ready examples instead of reading documentation.