Affiliate Disclosure: This post contains affiliate links. Read more →
Midjourney has the best AI image quality available in 2026 — and one of the steepest learning curves of any AI tool. It runs through Discord and has a command-based interface that feels unintuitive at first. After about an hour of practice, it becomes natural.
This guide takes you from never having used Midjourney to generating your first high-quality images and understanding the parameters that give you real control over the output.
TL;DR: Midjourney runs in Discord (or on the web at midjourney.com). Use /imagine + a text prompt to generate images. Learn the –ar (aspect ratio), –v (version), and –style parameters first — they give you the most control. Start with descriptive prompts, add style references, and iterate from there.
Table of Contents
- What You Need Before Starting
- Step 1: Join Discord and Midjourney
- Step 2: Subscribe to a Plan
- Step 3: Generate Your First Image
- Step 4: Upscale and Vary
- Step 5: Write Better Prompts
- Step 6: Use Key Parameters
- Step 7: Set Up a Private Server
- Tips and Tricks
- Common Mistakes to Avoid
- Murphy’s Take
- FAQ
- Sources
What You Need Before Starting
- A Discord account (free at discord.com)
- A Midjourney subscription (starting at $10/month — no free tier in 2026)
- Basic idea of what kind of images you want to create
If you prefer to skip Discord entirely, the Midjourney web app at midjourney.com now supports most features including image generation. This guide covers both methods.
Step 1: Join Discord and the Midjourney Server
- Create a Discord account at discord.com if you don’t have one
- Visit midjourney.com and click “Join the Beta”
- You’ll be redirected to the Midjourney Discord server
- Accept the invite and join
The Midjourney Discord server is large and public. You’ll see channels like #general, #newbies-1 through #newbies-10, and numerous topic channels. Start in one of the #newbies channels for your first images.
Alternative: Go directly to midjourney.com/app after subscribing to use the web interface instead of Discord.
Step 2: Subscribe to a Plan
- In Discord, type
/subscribein any channel - Click the link provided and choose a plan
- Or go to midjourney.com/account to subscribe directly
Plans start at $10/month (Basic, ~200 images). For regular use, the Standard plan at $30/month includes unlimited Relaxed mode generations, which is sufficient for most users.
Step 3: Generate Your First Image
In Discord, go to any #newbies channel (or a channel in a server where Midjourney bot is added) and type:
/imagine prompt: a red panda sitting on a mossy rock, forest background, golden hour lighting, photorealistic
Press Enter. Midjourney’s bot processes your prompt and returns four image variations in about 60 seconds.
On the web app: Type your prompt directly in the search/generate bar at midjourney.com/app and click the generate button.
Step 4: Upscale and Vary
After your four images appear, you’ll see buttons below them:
- U1, U2, U3, U4 — Upscale the corresponding image to full resolution
- V1, V2, V3, V4 — Generate four variations based on the corresponding image
- 🔄 (re-roll) — Generate four entirely new images from the same prompt
Upscaling creates a full-resolution version of the image for download. Use U1-U4 when one of the four images is close to what you want.
Varying generates alternatives that share the composition and style of the selected image but differ in details. Use V1-V4 when the direction is right but you want to explore options.
Step 5: Write Better Prompts
Good prompts have three components:
1. Subject — what’s in the image (“a medieval blacksmith”)
2. Style/aesthetic — how it should look (“oil painting, detailed brushwork, warm tones”)
3. Technical specs — camera, lighting, composition (“wide angle, dramatic side lighting, shallow depth of field”)
Examples:
Weak prompt: a cat
Strong prompt: a black cat sitting on a windowsill, rainy day outside, soft natural light, shallow depth of field, film photography style
Style keywords that consistently work:
– Photorealistic, cinematic lighting, shot on Hasselblad
– Watercolor, ink wash, gouache painting
– 3D render, octane render, unreal engine
– Concept art, illustration, storybook style
– Vintage photograph, film grain, kodachrome
Step 6: Use Key Parameters
Parameters are added at the end of your prompt with --. The most important ones:
| Parameter | Example | Effect |
|---|---|---|
--ar |
--ar 16:9 |
Aspect ratio (16:9 for widescreen, 9:16 for vertical, 1:1 for square) |
--v |
--v 6.1 |
Model version (6.1 is current default) |
--style |
--style raw |
raw = more photographic, less stylized |
--q |
--q 2 |
Quality (0.25, 0.5, 1, 2) — higher uses more GPU |
--chaos |
--chaos 30 |
Variation in the 4 results (0-100, higher = more diverse) |
--no |
--no text |
Negative prompt — exclude elements |
--sref |
--sref [url] |
Style reference from an image URL |
Example with parameters:
/imagine prompt: a Japanese tea garden at sunrise, misty, tranquil --ar 16:9 --style raw --q 2
Step 7: Set Up a Private Server
By default, all images you generate in Midjourney’s public channels are visible to everyone in the gallery. To keep your images private:
- Create your own Discord server (free)
- Go to midjourney.com → Manage subscription → Add to Server
- Add the Midjourney Bot to your personal server
- Use /imagine in your private server — images are not shown in the public gallery
Note: Stealth Mode (images hidden from the gallery on the website) requires the Pro plan ($60/month). A private server prevents your images from appearing in the community Discord, but they may still appear in your Midjourney profile gallery unless on a Stealth plan.
Tips and Tricks
- Use image references: Drag an image into Discord or paste an image URL at the start of your prompt to use it as a visual reference
- Try /blend: Combine 2-5 images to create a mashup of their styles and compositions
- Use Vary Region for fixes: After upscaling, click “Vary (Region)” to select and fix specific parts of the image
- Save useful prompts: Keep a notes document of prompts that work well for reuse and iteration
- Use multi-prompts: Separate prompt elements with
::and add weights (e.g.,forest ::2 castle ::1gives forest more emphasis)
Common Mistakes to Avoid
Writing too-complex prompts: More isn’t better. A 200-word prompt often produces worse results than a focused 20-word prompt. Start simple, add one element at a time.
Ignoring aspect ratio: Default is 1:1 square. If you’re creating images for social media, blogs, or specific layouts, set --ar first. Wrong aspect ratio means cropping, which changes composition.
Expecting perfect results first try: Midjourney requires iteration. Generate, vary the best option, adjust the prompt, repeat. Three rounds of iteration usually produces significantly better results than the first generation.
Murphy’s Take
The learning curve on Midjourney is real — the first session is confusing and the results are often not what you pictured. The second session is better. By the third or fourth session, you start to develop intuition for what the model responds to.
The thing I tell people starting out: don’t try to describe exactly what you want in every detail. Instead, describe the feeling and aesthetic of the image — cinematic, melancholic, ethereal, brutal — and let the model interpret it. The results are usually more interesting than a literal description.
Midjourney is a creative tool, not a search engine for images. The more you treat it as a creative collaborator rather than a request fulfillment machine, the better the results.
FAQ
Q: Do I need Discord to use Midjourney?
A: Not anymore. The Midjourney web app at midjourney.com/app supports image generation, upscaling, and most features. Discord was the original interface and many experienced users still prefer it, but new users can start entirely with the web app.
Q: How long does Midjourney take to generate images?
A: Fast mode generates four images in approximately 15-60 seconds depending on complexity and server load. Relaxed mode (available on Standard plan and above) can take 1-5 minutes when the server is busy.
Q: What’s the best aspect ratio for blog images?
A: Use --ar 16:9 for standard horizontal blog header images (1200×675px equivalent), or --ar 3:2 for a slightly more square format that works well in article headers. After upscaling, Midjourney images can be exported and resized to your exact pixel dimensions in any image editor.