Skip to content
Better AI Prompts

Midjourney Prompt Generator

Midjourney does not want a brief. It wants a dense line of description — subject, setting, light, lens, mood, medium — followed by the flags that control size and exclusions. Say what you picture in ordinary language and the generator writes it in that form, including the details photographers name and the rest of us only notice.

  • Free to use
  • Free account required
  • Your prompt stays yours
Characters used: 0/1000
Try one:

Only the description is required. Everything else sharpens the result.

Your prompt will appear here

Describe the job in your own words. We write the prompt, structured for the model you picked.

How it works

  • 1

    Description, not instructions

    No role, no objective, no output format. Image models read a list of qualities, so that is what comes back — one line, comma separated.

  • 2

    The vocabulary you are missing

    Golden hour, shallow depth of field, soft key light, 35mm. Naming these is most of what separates a striking render from a flat one.

  • 3

    Flags included

    Aspect ratio arrives as --ar, and anything you want kept out of frame as --no, already appended in the right place.

When to use it

  • Knowing the picture, not the words

    You can see it clearly and still not know that the look you mean is called rim lighting.

  • Keeping a series consistent

    Generate one prompt, then vary the subject while the lighting and lens stay fixed.

  • Excluding what keeps appearing

    Text, watermarks, extra hands. Say so once and it becomes a flag rather than a hope.

Frequently asked questions

  • Does this work for DALL-E or Stable Diffusion?

    The description part transfers well; the flags do not, since --ar and --no are Midjourney's. Remove them and the rest is still a good image prompt.

  • Can I upload a picture instead of describing one?

    Not yet — that is Image to Prompt, which is still being built. This page starts from your words.

  • Which version of Midjourney is this for?

    Anything recent. The flags used here have been stable across versions; niche parameters like --style raw you can add yourself.

  • Why is there no prompt score?

    There is one, but read it lightly. The seven measures were built for instruction prompts, and an image descriptor deliberately has no role or output format to score.

Related free tools