Setup and Getting Started
Midjourney is the leading AI art generator. Getting started takes just a few minutes.
Step 1: Join Discord
Midjourney runs entirely on Discord. If you don't have Discord, download it (free) from discord.com and create an account.
Step 2: Join Midjourney Server
Go to midjourney.com and click "Sign In" → "Join the Beta." This adds you to the official Midjourney Discord server.
Step 3: Subscribe
Use the `/subscribe` command in the Discord or visit midjourney.com/account to choose a plan:
| Plan | Price | Fast GPU Hours | Best For |
|---|---|---|---|
| Basic | $8/mo | 3.3 hours/month | Casual users, hobbyists |
| Standard | $16/mo | 15 hours/month | Regular creators, portfolio building |
| Pro | $30/mo | 30 hours/month | Professional artists, commercial use |
Step 4: Start Creating
You can now use the `/imagine` command in any Midjourney channel. Try a simple prompt first to get comfortable.
The /imagine Command
The core of Midjourney is the `/imagine` command. It takes your text description and generates an image.
Basic Format
/imagine [your detailed prompt]
Simple Example
/imagine A serene mountain field at sunset, oil painting style, warm golden light
Midjourney generates 4 image variations. You can then:
- U1-U4: Upscale one of the images (make it larger and add detail)
- V1-V4: Generate variations of one image (similar but different)
- Refresh: Re-generate all 4 with the same prompt
Prompt Structure Basics
The quality of your output depends on prompt quality. Follow this structure for best results:
Formula: Subject + Medium + Style + Mood + Details
Example broken down:
Subject: A young woman with flowing red hair
Medium: Digital painting
Style: Cyberpunk, neon lights
Mood: Determined, serious expression
Details: Holographic interfaces, dark background, 4k, ultra detailed
Full prompt: A young woman with flowing red hair, digital painting, cyberpunk style with neon lights, determined expression, holographic interfaces surrounding her, dark moody background, 4k ultra detailed --v 5.2
Better Prompts Checklist
Include:
- Clear subject matter (what's the main focus?)
- Artistic medium (oil painting, 3D render, photography, digital art)
- Style references (cyberpunk, surreal, minimalist, baroque)
- Mood and atmosphere (dark, bright, peaceful, chaotic)
- Composition hints (wide angle, close-up, bird's eye view)
- Lighting description (golden hour, neon, dramatic shadows)
- Quality notes (4k, highly detailed, photorealistic)
Avoid:
- Vague descriptions ("a nice picture")
- Too many conflicting styles
- Negation in prompts (instead of "not ugly," say "beautiful")
- Overly long prompts (150-200 words is ideal)
Style Keywords and Inspiration
Specific style keywords dramatically improve results. Use these as inspiration:
Artistic Mediums
- Oil painting, watercolor, acrylic, pen and ink
- Digital painting, vector art, pixel art
- Photography, cinematic, photorealistic
- 3D render, CGI, ray tracing
- Illustration, comic book, graphic novel
Art Styles
- Classic: Renaissance, Baroque, Impressionism, Art Nouveau
- Modern: Surrealism, Cubism, Abstract, Minimalism
- Contemporary: Cyberpunk, Steampunk, Vaporwave, Art Deco
- Pop Culture: Comic, anime, manga, cartoon, Studio Ghibli style
Mood Keywords
- Dark and moody, bright and cheerful, ethereal, dreamlike
- Chaotic, peaceful, energetic, melancholic
- Mysterious, epic, intimate, vast
Artist References
Mentioning artists gives Midjourney style direction:
/imagine A fantasy castle, in the style of Ralph McQuarrie and Studio Ghibli, oil painting, atmospheric lighting --v 5.2
Good artists to reference: Beksinski, Greg Rutkowski, Frank Frazetta, Artgerm, Simon Stalenhag, Concept art from movies
Aspect Ratios and Dimensions
Control image proportions with the `--ar` parameter:
Common Aspect Ratios:
--ar 1:1Square (Instagram, profile pictures)--ar 16:9Widescreen (YouTube thumbnails, desktop)--ar 9:16Vertical (mobile, TikTok, Instagram Stories)--ar 3:2Standard scene (photo prints)--ar 2:3Portrait (portrait photos)--ar 4:3Classic TV (presentations)
Example:
/imagine A sci-fi spaceship, sleek design, nebula background --ar 16:9 --v 5.2
Variations, Upscaling, and Remixing
Variations (V1-V4)
Creates new images similar to the selected image. Good for exploring different interpretations of your prompt.
Use when: You like the direction but want to see alternatives
Upscaling (U1-U4)
Enlarges the image and adds detail. Makes 512x512 into 1024x1024 (or higher with `--upbeta`).
Use when: You've found an image you love and want final quality
Remix Mode
Allows you to modify your prompt while keeping the composition. Enable with `/settings`, toggle Remix mode on, then modify your prompt when creating variations.
Example:
Original: "A dragon, fantasy style, golden"
Remix: "A dragon, fantasy style, blue and purple ice magic"
Advanced Parameters
Fine-tune results with these powerful parameters:
| Parameter | What It Does | Example |
|---|---|---|
--v 5.2 |
Model version (5.2 is latest) | Use for best results |
--style raw |
Less stylized, more literal | Use for photorealism |
--chaos 50 |
Randomness (0-100) | Higher = more variety |
--quality 2 |
Render quality (0.25, 0.5, 1, 2) | Higher = more detail |
--niji 5 |
Anime style model | Use for anime-style images |
Negative Prompts
Specify what you DON'T want using `--no`:
/imagine A beautiful world --no blurry, distorted, low quality --v 5.2
Pro Tips and Tricks
1. Test Simple Prompts First
Before crafting complex prompts, test basic versions to understand how Midjourney interprets concepts.
2. Use Image References
Upload an image to Discord and reference it in your prompt:
/imagine [image URL] but as a cyberpunk character, detailed --v 5.2
This guides the AI toward similar compositions or styles.
3. Iterate Strategically
Don't just generate randomly. Follow this workflow:
- Generate 4 initial images with a base prompt
- Pick your favorite
- Create variations to explore it further
- Once you love one, upscale it
4. Join Community Channels
The Midjourney Discord has public channels where you can see what others create. Observe successful prompts and learn from them.
5. Document Your Successful Prompts
Keep a running document of prompts that worked well. This builds your personal library of style keywords and structures.
6. Experiment with Weights
Control emphasis on specific parts of your prompt using `::` and weights:
/imagine A beautiful woman::3 in cyberpunk clothing::2, neon background::1 --v 5.2
Higher weight = more influence on that part
7. Use Blend for Mashups
Combine multiple images with `/blend`:
- Upload 2-5 images
- Choose aspect ratio
- Midjourney creates a new image blending them
8. High-Res Upscaling
Use `--upbeta` for 2x resolution upscaling without losing quality:
/imagine [prompt] --upbeta