To make an AI image, open Morphic, describe what you want to see in a prompt, name a style and an aspect ratio, then generate and refine the result. The prompt does the work: name the subject, the lighting, and the look, and the AI image generator turns those words into a finished picture.
The same idea, generated from a vague prompt and from a directed one.
Steps at a glance
- Describe the subject clearly
- Add style, lighting, and detail
- Choose the aspect ratio
- Generate and refine
- Upscale and export
Make your image step by step
1.
Describe the subject clearly
Start with what the image is of, in plain language. Name the main subject and what it is doing before anything else, because a vague subject is the fastest way to a generic result. "A coffee cup" leaves everything open; "a steaming ceramic coffee cup on a wooden table by a window" gives the model a scene to build. Get the subject and setting right first, then layer style and mood on top in the next step rather than cramming everything into one rushed sentence.
2.
Add style, lighting, and detail
Now direct how it looks. Name the style (photorealistic, illustration, 3D render, watercolour), the lighting (golden hour, soft studio, hard midday sun), and any detail that matters. Lighting is the single biggest lever on how real or how striking an image feels, so name a source and direction rather than leaving it flat. Keep each instruction short and specific, since a prompt that asks for ten things at once tends to deliver none of them cleanly.
3.
Choose the aspect ratio
Set the shape before you generate, because composition is decided at generation time and cropping afterwards throws away resolution. Pick a square for an Instagram post, a tall frame for a phone wallpaper or a Story, a wide frame for a banner or thumbnail. Matching the ratio to where the image is going means the subject is framed for that space from the start instead of awkwardly cropped later.
4.
Generate and refine
Generate, then treat the first result as a draft. If it is close but not right, change one thing, the light, the angle, a single detail, and regenerate rather than rewriting the whole prompt and losing what worked. You can also edit the result directly, adjusting part of the image or extending the frame. Generating a few variations and keeping the best is normal and costs only a moment.
5.
Upscale and export
When the image is right, upscale it for a larger, sharper version if you need print or full-screen resolution, then export. Upscaling and enhancement are paid features, and the free tier keeps upscale off, so a paid plan is what you want for high-resolution output. For anything published rather than tested, export from a paid plan to avoid a watermark.
Detailed prompt versus vague prompt
The prompt is the whole game. Here is the difference a directed prompt makes, with the same model.
| Vague prompt | Directed prompt | |
|---|---|---|
| Example | a woman in a cafe | a woman reading in a sunlit cafe, warm window light, film photo |
| Subject | Generic, unpredictable | Specific and consistent |
| Lighting | Flat, whatever the model picks | Named source and mood |
| Result | Looks generated | Looks intentional |
To make images of a specific look repeatedly, the text-to-image generator and the broader AI image generator are where these prompt patterns live. For a stylised take on a photo you already have, see how to turn a photo into AI art.
Put it together
Making an AI image is mostly prompt craft: name the subject and setting first, then direct the style and lighting, and set the aspect ratio before you generate. Treat the first result as a draft, change one thing at a time, and upscale only when the composition is right. Get the prompt specific and the model does the rest, turning a sentence into a finished picture in seconds. If you are making several images that need to match, keep the same style words and references across all of them, since a consistent prompt recipe is what turns one good image into a coherent set rather than a lucky one-off.