Magai

Free, no account, nothing saved

Free AI image prompt generator

Type the picture in your head, in a few words. Get back the prompt Midjourney, ChatGPT, Gemini, Flux, Stable Diffusion or Ideogram actually reads, with every part labeled so you learn what made it work.

Which image generator is the prompt for?

Midjourney reads: Comma-separated phrases, most important first, with parameters at the end.

Try

How to use the image prompt enhancer

  1. 1

    Say the idea plainly

    A few words is enough: who or what, and what is happening. The tool adds the rest.

  2. 2

    Pick the generator

    Each one reads prompts differently. The tool writes for the one you choose, in its own format.

  3. 3

    Copy, paste, generate

    Paste it as it is, or learn from the labeled parts and write the next one yourself.

What makes an AI image prompt work

A generator draws exactly what it is told and guesses the rest, and its guesses are the average of everything it has seen. Every part a prompt leaves out is filled with the most ordinary choice. So a good prompt is not longer for its own sake: it makes the choices a photographer or an illustrator would, one part at a time.

“an old lighthouse keeper making tea in a storm”

9 words

An elderly lighthouse keeper in a cable-knit wool sweater pours tea from a dented enamel pot. He sits in the cramped lantern room at dusk while rain streaks the curved glass behind him. Medium shot from a slightly low angle on a 35mm lens, the great lens softly out of focus. A warm tungsten lamp lights his face; cold blue storm light fills the windows. A cinematic film still with fine grain. Amber, slate blue and weathered brass. Quiet and content, a small warm room against a big storm. Steam curls from the cup and water beads on the brass fittings.

101 words, every one making a choice. Point at a part to see it alone.

  • Subject

    Who or what the image is of, specific enough to picture: age, material, pose, action.

  • Setting

    Where and when. A place and a time of day do more than any adjective.

  • Camera and framing

    Where the camera is and what it sees: shot size, angle, lens, depth of field.

  • Lighting

    The single biggest lever on quality. Name the source, its direction and its color.

  • Style and medium

    The medium, the rendering and the era: film stock, brushwork, render engine.

  • Color

    A palette, named. Two or three colors hold an image together.

  • Mood

    The feeling, said once. Generators copy emotion from the words around it.

  • Detail

    Texture and finish: the small true things that make it look made, not generated.

Built into Magai

In Magai, every prompt gets this treatment

This page writes image prompts. Inside Magai, enhancing a prompt is one click on the wand in the box you already type in, and it works on anything: an email, a code review, a research question, a plan. It rewrites what you typed into what an expert would have asked, so the model you picked answers the question you meant.

Any task
Writing, analysis, code, research, strategy and images. The enhancer adds the context, constraints and format the task needs.
Any model
Claude, ChatGPT, Gemini, Grok and dozens more. The better prompt goes to whichever one you choose.
Kept for next time
Save the prompts that work to your prompt library and run them again with one click.

The prompt library and enhancer · Everything else in Magai

What you typed

write an email telling a client the project is late
Claude
One click on the wand

What Claude receives

Write a short email to a client explaining that their website launch will move from March 3 to March 17. Open with the new date, give the reason in one sentence (the payment integration failed testing), say what we are doing about it and what they will see next week. Warm and accountable, no over-apologizing, under 150 words, with a subject line.
Claude

One idea, written for six generators

The same idea, "an old lighthouse keeper making tea in a storm", as each generator wants to read it. The words barely change; the shape does. The enhancer writes this shape for you.

Midjourney prompts

Midjourney weighs the start of a prompt most, and reads phrases better than sentences. Lead with the subject, keep it to one image, and put the shape and finish in parameters rather than words.

  • Put the thing you care about most in the first five words.
  • Use --ar for the shape, never the words “wide” or “vertical”.
  • Add --style raw for photographs; leave it off for illustration.

Midjourney prompt

elderly lighthouse keeper pouring tea, cable-knit sweater, cramped lantern room at dusk, rain on curved glass, medium shot, low angle, 35mm, shallow depth of field, warm tungsten lamp against cold blue storm light, cinematic film still, fine grain, quiet, content --ar 3:2 --style raw

ChatGPT image prompts

ChatGPT's image model follows instructions closely and reads whole sentences, so write it the way you would brief an illustrator. It is the best of the six at doing several specific things at once.

  • Say the shape in words: “a wide 3:2 frame”.
  • Put any words that must appear in the image in quotes.
  • Ask for one change at a time when you edit a result.

ChatGPT prompt

Create a photograph of an elderly lighthouse keeper in a cable-knit sweater pouring tea from a dented enamel pot. He is in the lantern room at dusk, rain streaking the glass behind him. Frame it as a wide 3:2 medium shot from a low angle. Light his face with a warm lamp and let cold blue storm light fill the windows. Make it look like a still from a film, with fine grain.

Gemini and Nano Banana prompts

Gemini's image model, the one people call Nano Banana, does best with a short scene described as a story, with the camera language of a photographer. It is especially good at keeping a person or product the same across edits.

  • Describe a scene, not a list of keywords.
  • Name the lens and the depth of field for realistic images.
  • For an edit, say what stays the same as clearly as what changes.

Gemini prompt

An elderly lighthouse keeper pours tea from a dented enamel pot, alone in the lantern room while a storm rolls in at dusk. Shot on a 35mm lens from a low angle, shallow depth of field, in a wide 3:2 frame. Warm lamplight on his face, cold blue light in the rain-streaked windows. The feeling is quiet contentment.

Flux prompts

Flux reads natural language with unusual precision and renders text well. It has no negative prompt, so describe what you want rather than what you do not, and be exact about light.

  • Say “clear sky”, never “no clouds”: there is no negative prompt.
  • Name the key light and the fill light separately.
  • Quote exact text and say where it sits.

Flux prompt

An elderly lighthouse keeper in a cable-knit sweater pours tea from an enamel pot, inside a lantern room at dusk with rain on the glass. Medium shot, low angle, 35mm lens, background softly blurred. Warm tungsten key light on his face, cool blue fill from the storm outside. Cinematic photograph, fine film grain. Steam rising from the cup, water beading on brass.

Stable Diffusion prompts

Stable Diffusion reads tags, weights the ones in parentheses, and leans on a negative prompt to steer away from common flaws. Quality terms at the front still help on most checkpoints.

  • Weight one or two tags, never all of them: (tag:1.2).
  • Keep a standard negative prompt and add to it per image.
  • Match the prompt style to the checkpoint: tags for anime models, phrases for photo models.

Stable Diffusion prompt

masterpiece, best quality, highly detailed, (elderly lighthouse keeper:1.2), pouring tea, wool sweater, lantern room, dusk, rain on glass, medium shot, low angle, 35mm, depth of field, (warm tungsten light:1.1), blue storm light, cinematic, film grain

Negative prompt: blurry, lowres, bad anatomy, extra fingers, deformed hands, watermark, text, oversaturated

Ideogram prompts

Ideogram is the one to reach for when the image has words in it: signs, posters, logos, cards. Quote the text exactly, and say where it goes and what the lettering looks like.

  • Put every word you want rendered in quotes.
  • Name the lettering: serif, script, stencil, hand-painted.
  • Keep the text short; a few words renders far better than a sentence.

Ideogram prompt

A vintage enamel sign reading “Keeper's Tea” hangs inside a lighthouse lantern room, in cream serif lettering on deep navy with a thin brass border. Behind it, an old keeper pours tea while rain streaks the glass at dusk. Warm lamplight, cold blue windows. Wide 3:2 frame, the sign sharp in the foreground.

Words that change the picture

Most of the lift in a good prompt comes from three kinds of word: where the light comes from, where the camera is, and what the image is made of. These are the ones generators know best.

Lighting

Golden hour
Low, warm sun just after sunrise or before sunset. Long shadows, flattering skin.
Blue hour
The cool twilight after sunset. City lights glow against a deep blue sky.
Rim light
Light from behind that outlines the subject and lifts it off the background.
Softbox
Large, diffused studio light. Even, clean, commercial.
Chiaroscuro
Strong light against deep shadow, as in old master paintings.
Overcast
Flat, shadowless daylight. Good for products and true color.

Camera and framing

Close-up
The face or object fills the frame. Intimate and detailed.
Wide shot
The whole subject and its surroundings. Tells you where you are.
Low angle
Camera below the subject, looking up. Makes it feel large or heroic.
Overhead
Straight down from above. The standard for food and flat lays.
85mm lens
The portrait lens: flattering faces, background melted away.
Shallow depth of field
Only the subject is sharp. Separates it from a busy background.

Style and medium

Film still
Looks like a frame from a movie: cinematic color, a story in progress.
Editorial photography
Magazine polish: styled, lit, and composed on purpose.
Watercolor
Soft washes, visible paper, colors bleeding at the edges.
Isometric
A 3D view with no perspective. Tidy diagrams and tiny worlds.
Claymation
Handmade clay figures with fingerprints and soft studio light.
Risograph
Grainy two or three color print with slightly misaligned layers.

AI image prompt ideas to start from

Pick one and it goes straight into the enhancer. Then change a word and run it again: that is how you learn what each part does.

Portraits

Products

Logos and text

Places

Just for fun

Questions about image prompts

Better prompts for every task, on every model

Magai puts every leading AI model and several image generators in one workspace, with one-click prompt enhancement in the box you type in. Save what works to your prompt library and reuse it anywhere.

Get started See the image models
30-day money-back guarantee on every plan.