Help Centre: HubPeople

Parameters go at the end of the prompt, after the description. Midjourney reads comma separated concepts rather than full sentences, so describe in fragments and control with flags.

Core settings and aesthetics

Parameter	What it does
--ar W:H	Defines the canvas shape. Use --ar 16:9 for desktop hero banners, --ar 9:16 for mobile views, and --ar 1:1 for profile avatars.
--stylize (--s), 0 to 1000	Controls the built in artistic finish. Low values (10 to 50) follow the text prompt literally, mid values (100 to 150) offer balanced photorealism, high values (300 and above) add heavy aesthetic embellishment.
--style raw	Disables Midjourney's automatic cinematic bias, producing clean, photorealistic camera behaviour and stricter prompt adherence. This is the single biggest lever against the "AI look".
--chaos (--c), 0 to 100	Sets how varied the four initial grid results are. 0 gives minor tweaks, 50 and above yields completely distinct scene compositions.
--weird (--w), 0 to 3000	Introduces surreal, offbeat visual elements. Keep at 0 for commercial web design.

In editor controls and retouching

Control	What it does
Pan (arrows)	Extends the canvas horizontally or vertically without altering the centre image. Ideal for stretching a landscape background into a wide banner.
Zoom out (1.5x, 2x, custom)	Pulls the virtual camera back from the subject to create negative space around a centred focal point.
Vary (Region), inpainting	Brush over part of an image and change just that part with a text prompt, for example recolouring an outfit or fixing a distorted hand.
Web editor	One workspace combining pan, zoom, inpainting and manual canvas resizing.
Retexture (style swap)	Replaces lighting, colour and art style while preserving the exact geometry and subject features.

Character and brand consistency

This is the group that matters most for a brand: it is how you keep the same face and the same look across every page of a site, rather than a different stranger on each one.

Parameter	What it does
--cref [image]	Character reference. Locks a specific face and facial structure across multiple prompt variations, so the same model appears site wide.
--cw [0 to 100]	Character weight, used alongside --cref. --cw 100 copies hair, face and clothing. --cw 0 locks only the face, so clothing and environment can change.
--sref [image]	Style reference. Copies visual tone, colour grading and lighting from a reference photo without copying its subjects.

Advanced prompting

Negative prompting

Filter out unwanted elements with --no. Useful for pushing away clichéd poses and the tells that make an image read as generated.

--no text, watermark, hand holding, CGI, plastic skin

Multi prompt weighting

Two colons assign relative importance to concepts, so the model knows what the picture is actually about. Here the couple matters four times as much as the crowd behind them.

senior couple::2 crowd::0.5

Text rendering

Quotation marks instruct the model to render legible text on surfaces in the image, for example a pub sign or a shopfront.

a traditional pub sign reading "The Coach and Horses"

Permutation prompts

Curly braces generate several variations from one command, running the jobs at the same time.

couple {laughing, talking} --ar 16:9

Image prompting

Placing an image URL before the text prompt uses that photo as visual guidance for the result.

A worked starting point

The workshop ran one brief through four engines. This was the Midjourney version, and it is a reasonable template to adapt: comma separated fragments grouped by subject, background and composition, with the flags last.

a candid, hyper-realistic photo of a sophisticated senior couple, aged 55-65, laughing genuinely together while seated outdoors at a traditional pub in Soho, London, sunny afternoon
Subjects: classy woman with elegant grey pixie cut, tailored navy blazer, gold necklace; distinguished man, short white hair, neat beard, dark green waxed cotton jacket, smiling
Background: traditional British pub facade, dark wood paneling, gold signwriting, red hanging flower baskets, blurred pedestrians, cobblestone street (bokeh)
Composition: dead center, significant negative space left/right, shot on Fujifilm Velvia 50, shallow DOF, golden hour, cinematic, high detail, photorealistic
--ar 16:9 --style raw --chaos 5 --stylize 150 --no holding hands

Dial in ratio and style first, then use chaos and negative prompts to fine tune. Naming a camera or a lens gives real control over the look. Stuck for a starting point? The Explore tab in Midjourney is worth a browse.