# Midjourney photorealistic prompts: a practical method

Canonical: https://romeoapps.app/en/larpgpt/midjourney-photorealistic-prompts/
Author: Roméo Gambino · Published: 2026-09-10 · Updated: 2026-09-10 · App: LarpGPT (https://romeoapps.app/en/larpgpt/)

How to write Midjourney prompts that read as photographs: a camera-first structure, the documented parameters that help, and how to iterate.

## What photorealistic actually means in a prompt

A photorealistic image is one that could plausibly have come out of a camera. That sounds obvious, but it changes how you write. You are not asking for realism as a quality; you are describing the conditions a photographer would have faced: a real place, a real time of day, a real light source, a lens at a certain distance from the subject. When those conditions are stated, the result tends to look recorded. When they are missing, the model fills them in with its own defaults, and defaults rarely look like a specific photograph.


## Why the word photorealistic is not enough

Adding photorealistic, ultra realistic or 8k to the end of a vague prompt is the most common approach and the least useful one. Those words describe an outcome without giving the model anything to build it from. A prompt that says an old fishing boat pulled up on grey shingle, overcast morning, eye-level view from three metres away already implies realism, because every part of it is something a camera could record. The adjectives can stay if you like them, but they are not what does the work.


## Start with a subject you could actually photograph

Describe the subject in physical terms: material, age, condition, size, what it is doing. A leather armchair with a worn left armrest is photographable. A luxurious chair is an opinion. Physical detail gives Midjourney surfaces to render, and surfaces are where realism is won or lost. Wear, dust, fingerprints, creases and small asymmetries are what make an object look like it exists, because real objects are never perfectly clean or perfectly symmetrical.


## Put it somewhere specific

A subject floating in no particular place reads as a render. Name the setting and let it constrain everything else: a narrow Lisbon street with tiled facades, a concrete car park at dusk, a kitchen with a single window over the sink. The setting decides what light is available, what surfaces reflect it and what can appear behind the subject. It also gives the image a scale, which is one of the quiet signals that separate a photograph from an illustration.


## Name the light and the moment together

Time of day and weather carry an entire lighting setup in a few words. Late afternoon sun implies a low, warm light and long shadows. An overcast sky implies soft, even light with gentle shadows. Night on a city street implies mixed artificial sources with different colours. Write the moment and the source together, for example "low evening sun from the left, long shadows across the pavement", and you have described most of what makes a picture feel photographic.


## Describe the camera position, not the camera brand

Where the camera stands matters more than which camera it is. Eye level, low angle, overhead, a few steps back, close enough to fill the frame: these are instructions about geometry, and geometry is something the model can follow. Lens language helps in the same way. A wide lens exaggerates depth and makes near objects large; a long lens compresses distance. A shallow depth of field isolates the subject. Naming a specific camera model adds little that these plain descriptions do not already say.


## How parameters work in Midjourney

Parameters are short instructions added after your description, and Midjourney's documentation is precise about the format. Parameters go at the end of the text prompt, with a space between your text and the two dashes, and without commas or full stops inside the parameters themselves. A prompt such as "weathered stone farmhouse at dawn, mist in the valley --ar 3:2" follows that rule. Text written after a parameter, or punctuation attached to it, is listed in the documentation as a formatting mistake.


## Set the aspect ratio before anything else

Midjourney images start as squares. The documentation lists --ar, or its long form --aspect, to change that, written as two whole numbers such as --ar 3:2 or --ar 9:16; decimals are not accepted. It also notes that extremely wide or tall ratios are experimental and that aspect ratio is not the same as final pixel dimensions. Choosing the ratio first matters for realism, because a photograph's framing is decided before the picture is taken, and a composition designed for a square rarely survives being stretched.


## Raw mode for a less stylised starting point

Midjourney documents a Raw Mode, set by adding --raw to the end of the prompt or by choosing it in the settings panel. According to the documentation, standard mode adds Midjourney's own creative touch, while Raw Mode turns off that automatic styling, so simple prompts often produce more photo-like images and detailed prompts give you more control over the final look. For photographic work it is a sensible default to test, and a reason to write your description carefully, since less is added for you.


## Use stylize as a dial, not a switch

The --stylize parameter, shortened to --s, controls how much artistic interpretation Midjourney applies. The documentation describes a default of 100 and a range from 0 to 1000 in the latest versions: lower values follow the prompt more closely, higher values give the model more freedom and may stray from your details. For realism, try a few lower values on the same prompt and compare them side by side rather than assuming the lowest number is always the most photographic.


## Leave things out with --no, sparingly

The --no parameter tells Midjourney what you do not want, followed by one element or a comma-separated list, for example --no text, logo. The documentation warns that each word after --no is read on its own, so a phrase like modern clothing is read as two separate exclusions. It suggests describing what you do want instead in that kind of case. For photographic prompts, a short exclusion list for text, watermarks or clutter is usually enough; a long one tends to create new problems.


## Iterate one change at a time

When a result is close but wrong, change a single element and generate again. Swap the light, keep everything else. Move the camera lower, keep everything else. This is slower per attempt and much faster overall, because you learn what each part of the prompt is responsible for. Rewriting the whole prompt after every result feels productive, but it makes it impossible to tell which change fixed the image and which one quietly broke something else.


## Check the details a photograph would get right

Before you use an image, look where generated pictures often fail: text on signs and labels, hands, reflections in glass and mirrors, the direction of shadows compared with the stated light source, and repeated patterns such as windows or tiles. A realistic image is not only one that looks good at a glance; it is one where the physics hold up when someone looks closely. If the image might be taken for a real photograph, say that it was generated.


## Where LarpGPT fits

LarpGPT is a library of photo prompts for travel, architecture, cars and lifestyle scenes, each written around subject, composition, lighting and lens. You copy a prompt, adapt the details, and paste it into Midjourney or any other image generator you already use, then add your own parameters. LarpGPT does not generate images itself, and what the generator produces varies from one attempt to the next.


## Questions

### How do I make Midjourney images look like real photos?

Describe the conditions a photographer would face: a physical subject, a specific place, a named light source and time of day, and a camera position. Words like photorealistic add little on their own; the described conditions do the work.

### Where do parameters go in a Midjourney prompt?

At the end, after your description, with a space before the two dashes and no punctuation inside the parameters, according to Midjourney's documentation.

### What does Raw Mode do?

Midjourney's documentation says it turns off the automatic creative styling of standard mode, so simple prompts often look more photo-like and detailed prompts give you more control. You enable it with --raw or in the settings.

### Should I set stylize to zero for realism?

Not automatically. The documented range runs from 0 to 1000 with a default of 100, and lower values follow the prompt more closely. Compare a few low values on the same prompt and keep the one that suits the image.

## Further reading

- Midjourney documentation: Parameter List (archived copy, September 2026) — https://web.archive.org/web/20260905023030/https://docs.midjourney.com/hc/en-us/articles/32859204029709-Parameter-List
- Midjourney documentation: Aspect Ratio (archived copy, May 2026) — https://web.archive.org/web/20260519054426/https://docs.midjourney.com/hc/en-us/articles/31894244298125-Aspect-Ratio
- Midjourney documentation: Raw Mode (archived copy, October 2025) — https://web.archive.org/web/20251016193559/https://docs.midjourney.com/hc/en-us/articles/32634113811853-Raw-Mode
- Midjourney documentation: Stylize (archived copy, August 2026) — https://web.archive.org/web/20260808042710/https://docs.midjourney.com/hc/en-us/articles/32196176868109-Stylize
- Midjourney documentation: No (archived copy, July 2026) — https://web.archive.org/web/20260731111101/https://docs.midjourney.com/hc/en-us/articles/32173351982093-No