Midjourney Prompt Generator

Mastering Midjourney prompts

Midjourney is a closed-source text-to-image diffusion model created by David Holz (co-founder of Leap Motion) and launched in July 2022. Until 2024 it ran exclusively inside a Discord bot via the /imagine command, before adding a Web app at midjourney.com. Among the big three text-to-image services, it is the one most associated with a signature painterly aesthetic — even minimal prompts come out cinematic and richly lit, which is great for moodboards but harder to tame for strict art direction.

Major versions go from v1 through v6 (current as of early 2024), plus the Niji family for anime/manga aesthetics. Each new version raises coherence and detail — v6 finally became reliable with text-in-image and hands.

Prompt syntax and parameters

A full prompt looks like /imagine prompt: a samurai cat in cherry blossoms --ar 16:9 --v 6 --style raw --stylize 100. The most useful parameters are --ar (aspect ratio: 16:9, 9:16, 1:1, 3:2, 4:3), --v (version), --niji (anime model), --stylize (0-1000, default 100, controls how artistically Midjourney interprets the prompt), --chaos (0-100, variety of the 4-image grid), --weird (0-3000, surreal mode), --seed (reproducibility), --no (negative, removes an element), --style raw (less Midjourney-aesthetic, more literal) and --tile (seamless pattern).

Image weights, style refs and permutations

Image weights via --iw 0.5 to --iw 3.0 balance how much a reference image counts versus the text. Style references use --sref <URL> to borrow the aesthetic of an image without copying its subject. Multi-prompts let you weight tokens directly: cat::2 mouse doubles the importance of cat. Permutations like {cat, dog, horse} on a skateboard spawn three variants at once.

Prompt anatomy that consistently works

A formula that ages well: subject + descriptors + style + lighting + camera lens + film stock + artist reference. Example: a samurai cat in cherry blossoms, golden hour, 85mm portrait lens, shot on Kodak Portra 400, by Greg Rutkowski and Studio Ghibli. Specifying lenses (35mm wide, 50mm natural, 85mm portrait), film stocks (Kodak Portra 400, Fuji Velvia, CineStill 800T) and named artists is the fastest way to pin down a look that is otherwise too generic.

Pricing and how it compares

Plans are Basic $10/mo (~3.3h of GPU), Standard $30 (~15h plus relaxed mode), Pro $60 (~30h plus stealth mode for private generations) and Mega $120. vs Stable Diffusion: closed and paid versus open-source and free, but Midjourney needs zero setup and ships a recognisable aesthetic by default. vs DALL·E 3: Midjourney is stronger on artistic rendering, DALL·E 3 wins on literal instruction following and on text inside the image.

FAQ

Is it better to omit --v? No — without it you get the default version of your account, and pinning --v 6 (or whatever is current) keeps your prompt reproducible later on.

Is 16:9 the right aspect ratio for mobile content? No, mobile prefers 9:16 vertical (TikTok, Reels, Shorts). 16:9 is for YouTube thumbnails, desktop wallpapers and presentations.

Can I sell images generated by Midjourney commercially? Yes on paid plans — the Terms grant full ownership of generated images to paying subscribers. The Pro plan adds stealth mode so your prompts and outputs are not visible in the public gallery.

Why are my photos too painterly? Midjourney's default style leans artistic. Add --style raw to dampen it, lower --stylize to 50, and explicitly add words like photograph, shot on Kodak Portra 400, 85mm lens.