How to Use Midjourney in 2026

What You Need Before You Start

Before diving in, here's what you need to know: Midjourney requires a paid subscription. There is no free trial. You'll need to subscribe before generating any images.

What you need:

• A Midjourney account — Sign up at midjourney.com
• A subscription plan — Starting at $10/month (Basic)
• Optionally, a Discord account — Required if you want to use Discord instead of the web interface

You can use Midjourney entirely through the web interface at midjourney.com. Discord is optional but offers some unique features like community channels and certain commands. Most beginners start on the web.

Choose a Plan

Midjourney plans are based on GPU time. Every image you generate uses GPU minutes. A typical 4-image grid takes roughly 1 minute of GPU time. Video generation costs more.

Speed modes:

• Fast Mode — Priority processing, typically 30–60 seconds. Uses your GPU hours.
• Relax Mode — Queued processing, 1–10 minutes. Free and unlimited (Standard+).
• Turbo Mode — Fastest available, ~15–30 seconds. Costs 2× GPU time.

Current plans (2026):

• Basic — $10/mo — ~3.3 hours Fast GPU time (~200 images). No Relax mode.
• Standard — $30/mo — 15 hours Fast GPU + unlimited Relax. Most popular.
• Pro — $60/mo — 30 hours Fast + unlimited Relax + Stealth Mode.
• Mega — $120/mo — 60 hours Fast + unlimited Relax + Stealth.

Tip: Annual billing saves ~20%. For most users, Standard is the sweet spot—unlimited Relax mode means you can generate endlessly when you're not in a rush.

Create Your Midjourney Account + Subscribe

Step A: Sign Up at midjourney.com

1. Visit midjourney.com
2. Click "Sign In" or "Sign Up"
3. You can create a standalone account or sign in with Discord/Google
4. Verify your email if prompted

Step B: Subscribe to a Plan

1. Once signed in, you'll be prompted to choose a plan
2. Select the plan that fits your needs (you can upgrade or downgrade later)
3. Enter payment details
4. You're ready to start generating!

Start Generating on the Web (midjourney.com)

The web interface is the most popular way to use Midjourney. Here's how to create your first image:

1. Go to the Create page — Click "Create" in the navigation or go directly to the creation interface
2. Find the Imagine bar — This is where you type your prompt
3. Click the settings icon (gear) — Adjust model version, aspect ratio, and other defaults
4. Enable Draft Mode (optional) — Toggle the Draft Mode button for faster, cheaper previews
5. Type your prompt and hit Enter — Or use Ctrl+Enter to quickly submit

Your first prompt — try this:

cozy reading nook in a small cabin, warm window light, wool blanket, coffee mug --ar 4:3 --stylize 150

After 30–60 seconds, you'll see a 4-image grid. Click any image to view it larger, then use the buttons below to upscale, create variations, or expand.

Start Generating in Discord

Discord is the original Midjourney interface and still offers unique features. Here's how to set it up:

Connect Discord to Your Account

1. Go to your account settings on midjourney.com
2. Link your Discord account
3. Join the Midjourney Discord server: discord.gg/midjourney

Using the /imagine Command

1. Go to any "newbie" or bot channel
2. Type /imagine and press Tab
3. Enter your prompt in the text field
4. Press Enter
5. First time? Accept the Terms of Service when prompted

Understanding the Grid + Buttons

When your image generates, you'll see a 4-image grid with buttons below:

• U1, U2, U3, U4 — Upscale that image to full resolution
• V1, V2, V3, V4 — Create variations of that image
• 🔄 (Reroll) — Generate 4 completely new images with the same prompt

Pro tip: Type /info to check your remaining GPU time and subscription status.

Prompting Basics

The prompt is everything. Here's what actually works:

Be clear, not long. Midjourney works best with concise, descriptive prompts. You don't need paragraphs. A focused sentence often beats a wall of text.

Short prompts work best. The model has limited attention. Front-load the most important details.

The Prompt Recipe

Think of prompts in building blocks:

• Subject — What is it? (a fox, a castle, a portrait)
• Details — Specifics (red fur, ancient stone, young woman)
• Environment — Where? (forest, mountaintop, studio)
• Style — Art style (oil painting, 3D render, photograph)
• Lighting — (golden hour, dramatic rim light, soft diffused)
• Camera/Mood — (close-up, wide angle, moody, ethereal)

Example using the recipe:

a red fox (subject) with fluffy winter coat (details) sitting in a snowy forest (environment), painted in watercolor style (style), soft morning light filtering through trees (lighting), peaceful serene mood (mood)

More Tips

• Use positive wording — Say what you want, not what you don't want. Instead of "no people," describe an empty scene.
• Numbers help — "three apples" is more reliable than "some apples"
• Avoid conflicting concepts — "minimalist baroque" confuses the model

Parameters Overview

Parameters let you control how Midjourney interprets your prompt. They go at the end of your prompt and start with -- (two dashes).

Syntax rules:

• Parameters always go at the end of your prompt
• Put a space before each --
• No punctuation inside parameters (no commas, periods)
• Multiple parameters can be chained: --ar 16:9 --stylize 250 --chaos 20

Example:

a mountain landscape at sunset --ar 16:9 --stylize 200 --v 7

Most-Used Parameters

--ar (Aspect Ratio)

Controls the width-to-height ratio of your image. Default is 1:1 (square).

• --ar 16:9 — Widescreen landscape
• --ar 9:16 — Vertical/portrait (great for phone wallpapers)
• --ar 4:3 — Classic photo ratio
• --ar 3:2 — Traditional photography
• --ar 21:9 — Ultra-wide cinematic

--stylize or --s (Artistic vs Literal)

Controls how much Midjourney applies its trained aesthetic. Range: 0–1000. Default: 100.

• Low (0–50) — More literal, follows prompt closely, less "beautified"
• Medium (100–250) — Balanced, default Midjourney look
• High (500–1000) — Very artistic, may interpret prompt loosely

--chaos or --c (Variety in Grid)

Controls how varied the 4 images in your grid are. Range: 0–100. Default: 0.

• 0 — All 4 images are similar variations
• 50 — Moderate variety
• 100 — Wildly different interpretations

--weird or --w (Experimental)

Adds unexpected, experimental elements. Range: 0–3000. Default: 0.

Use sparingly. High values can produce surreal, unsettling, or avant-garde results.

--raw (Reduces Auto-Styling)

Tells Midjourney to apply less automatic "beautification." No value needed—just add --raw.

Useful when you want more control or a more literal interpretation of your prompt.

--quality or --q (Detail/Texture)

Controls rendering quality and detail level. Values: 0.25, 0.5, 1 (some versions support 2). Default: 1.

• --q 0.5 — Faster, less detail (good for quick tests)
• --q 1 — Standard quality
• --q 2 — More detail/texture (costs more GPU time)

--seed (Lock Random Start)

Every image starts from a random seed. Lock it to get reproducible results. Range: 0–4294967295.

Same prompt + same seed = similar image. Useful for iterating on a specific composition.

--no (Exclude List)

Tells Midjourney to try to avoid certain elements. Comma-separated.

a serene lake --no people, boats, buildings

Note: --no isn't perfect. It reduces likelihood, doesn't guarantee exclusion.

--tile (Seamless Patterns)

Creates images that tile seamlessly. Great for textures, fabrics, wallpapers.

geometric art deco pattern, gold and navy --tile

--repeat or --r (Multiple Grids)

Generates multiple 4-image grids from one prompt. Range: 2–40.

fantasy landscape --r 4

This generates 4 separate grids (16 images total). Great for exploring variations.

--version or --v (Model Version)

Selects which Midjourney model to use. Default is V7 (latest).

• --v 7 — Latest version, best overall quality
• --v 6.1 — Previous stable version
• --v 5.2 — Older version, different aesthetic

--niji (Anime-Focused Model)

Switches to Niji model, optimized for anime and illustration styles.

a magical girl with flowing hair, sparkles --niji

Using Your Own Images

Midjourney can use your own images as references. There are three main ways to do this:

A) Image Prompts

Paste an image URL at the start of your prompt. Midjourney will use it as a visual reference.

https://example.com/your-image.jpg a portrait in similar style --iw 1.5

--iw (image weight) — Controls how much influence the image has. Range: 0–3. Default: 1.

• --iw 0.5 — Light reference, more freedom
• --iw 2 — Strong reference, closely follows image

B) Style Reference (--sref)

Use an image to define the style (not the content) of your generation.

a medieval castle --sref https://example.com/style-image.jpg --sw 100

• --sref — The style reference image URL
• --sw (style weight) — How strongly to apply the style. Range: 0–1000. Default: 100.
• --sv — Style version (for different style interpretations)

C) Omni Reference (--oref)

New in V7: Omni Reference replaced Character Reference. It lets you reference specific characters, objects, or subjects across generations.

a warrior in battle --oref https://example.com/my-character.jpg

Limitations: Works best with clear, well-lit reference images. Complex poses or unusual angles may not transfer perfectly.

Web vs Discord for Reference Images

• Web: Upload images directly or paste URLs. Easier for most users.
• Discord: Attach images to your message, or paste URLs. The image must be publicly accessible.

Multi-Prompts and Weights

Use :: to split your prompt into separate concepts. This helps when you want Midjourney to treat parts of your prompt independently.

Basic Multi-Prompt

hot dog

↑ This might give you a hot dog (food) OR a hot dog (warm puppy).

hot:: dog

↑ This separates "hot" and "dog" as distinct concepts.

Adding Weights

Add numbers after :: to weight each section:

cyberpunk city::2 neon lights::1 rain::0.5

Higher weights = more emphasis on that part of the prompt.

Negative Weights

Use negative values to de-emphasize concepts:

a forest scene::1 people::-0.5

(This is an alternative to --no)

Iteration Workflow

Great images rarely come from a single prompt. Here's a 6-step iteration loop:

1. Start simple — Write a basic prompt. Don't overthink it.

   a cozy cabin in the woods

2. Choose your favorite — Look at the 4-image grid. Pick the one with the best composition or vibe.
3. Refine with Remix — Enable Remix mode (in settings), then hit V1-V4. You can modify the prompt while keeping the composition.
4. Fix with Vary Region / Erase — Select specific areas to regenerate. Fix hands, faces, or unwanted elements.
5. Expand with Pan / Zoom — Use Pan (arrows) to extend the image in any direction. Use Zoom Out to add more context around your subject.
6. Finalize quality — Upscale your final image. Use Subtle or Creative upscale for different effects.

Key insight: Iteration beats prompt engineering. A mediocre prompt + good iteration beats a perfect prompt every time.

Draft Mode

Draft Mode generates faster previews at half the GPU cost. Perfect for exploring ideas before committing.

• Enable on web: Toggle the "Draft Mode" button in the create interface
• Enable via parameter: Add --draft to your prompt
• Finalize: Use the "Enhance" button to render at full quality

When to use Draft Mode:

• Exploring composition ideas
• Testing prompts quickly
• When you're not sure what you want yet

When to skip it:

• You need final quality immediately
• Working on detailed textures or fine details

Video in Midjourney

Midjourney can now generate short videos! Starting at 5 seconds, video generation uses more GPU time than images.

Video Parameters

• --motion — Enables video generation
• --raw — Less stylization (works in video too)
• --loop — Creates a seamless looping video
• --end — Specifies an end frame (use with image reference)
• --bs — Adjusts motion intensity / blur style

Example video prompt:

a candle flame flickering in darkness --motion --loop --ar 9:16

Note: Video costs more GPU time than images. Check your /infobefore generating many videos.

Essential Discord Commands

If you use Discord, here are the commands you'll need:

Core Commands

• /imagine — Generate images from a prompt
• /settings — Open your settings panel
• /info — Check subscription status and GPU time remaining

Utility Commands

• /describe — Upload an image, get prompt suggestions
• /blend — Combine 2–5 images into one (no text prompt needed)
• /show — Retrieve a job by its ID
• /shorten — Analyze your prompt, suggest a shorter version

Preference Commands

• /prefer suffix — Auto-add parameters to all your prompts
• /prefer option set — Create custom parameter shortcuts

Privacy & Speed

• /stealth — Enable Stealth Mode (Pro/Mega only)
• /public — Disable Stealth Mode
• /fast — Switch to Fast Mode
• /relax — Switch to Relax Mode
• /turbo — Switch to Turbo Mode

Privacy, Visibility & Organization

Stealth Mode

Stealth Mode (Pro and Mega plans only) hides your images from the public Midjourney gallery. Your work won't appear on the Explore page or be visible to other users.

Important: Stealth Mode only affects the public gallery. Your images are still stored on Midjourney's servers and subject to their terms of service.

Private Discord Server

You can invite the Midjourney bot to your own private Discord server. Your generations are still processed by Midjourney (not truly private), but other Discord users won't see your prompts in public channels.

The Trash Warning

When you "trash" an image on Midjourney, it's marked for deletion but not immediately removed. Be mindful of what you generate, even if you delete it.

The Organize Page

Use the Organize page on midjourney.com to manage your images. You can:

• Create folders and collections
• Search your generation history
• Download images in bulk
• Delete old generations

Troubleshooting

Commands Not Showing in Discord

• Make sure you're in a channel where the Midjourney bot has permissions
• Try typing / and waiting for the command list to load
• If the bot isn't visible, you may need to join a newbie channel first
• Check that your Discord account is linked to your Midjourney subscription

Image URLs Not Working

• URLs must be direct image links (ending in .jpg, .png, .webp, etc.)
• The image must be publicly accessible (not behind a login)
• Google Drive, Dropbox, etc. need special share link formatting
• Try uploading the image directly to Discord or the web interface instead

Results Don't Match My Prompt

• Front-load important details — Midjourney pays more attention to the beginning
• Reduce stylize — Try --s 50 for more literal results
• Use --raw — Reduces automatic beautification
• Be more specific — "a dog" is vague; "a golden retriever puppy, close-up portrait" is clear
• Remove conflicting concepts — "photorealistic watercolor painting" confuses the model

Starter Prompt Pack

Copy these prompts to get started. Each demonstrates different techniques:

Photoreal Product Photo

a sleek wireless earbud case on a marble surface, soft studio lighting, product photography, shallow depth of field, minimalist background --ar 4:5 --stylize 80

Children's Book Illustration

a friendly dragon teaching a little girl to fly, whimsical children's book illustration, soft watercolors, warm palette, cozy and magical atmosphere --ar 3:2 --stylize 200 --niji

Pattern for Fabric (Seamless Tile)

Japanese wave pattern with koi fish, traditional ukiyo-e style, indigo and white, seamless textile design --tile --ar 1:1

Testing with a Seed

Use a fixed seed to test how different parameters affect the same base image:

a mountain landscape at golden hour --seed 12345 --stylize 100

Now try changing --stylize to 50, 200, and 500 while keeping the same seed.

Rules and Safety

Midjourney has community guidelines. Breaking them can result in warnings or bans.

Be Respectful

• No harassment, hate speech, or discrimination
• Don't generate images of real people without consent (especially harmful/sexual content)
• Respect copyright and intellectual property

Content Guidelines

• No explicit adult content (Midjourney has automatic filters)
• No violence or gore beyond what's appropriate for a PG-13 movie
• No content depicting minors in inappropriate situations
• Avoid generating misleading deepfakes or disinformation

Commercial Use

All paid subscribers can use their generated images commercially. You own the rights to your creations. However, you cannot copyright AI-generated images in most jurisdictions.

Recommended Beginner Path

Here's a structured learning path for getting good at Midjourney:

1. Start on the web — Easier interface, no Discord complexity. Generate 10-20 images to get comfortable with the basics.
2. Learn 6 key parameters — Master these first:
   • --ar (aspect ratio)
   • --stylize (artistic vs literal)
   • --chaos (variety)
   • --no (exclusions)
   • --seed (reproducibility)
   • --v (model version)
3. Use Draft Mode — Save GPU time while exploring. Get comfortable with the draft → enhance workflow.
4. Try Style Reference — Find an image with a style you love. Use --srefto apply it to your prompts.
5. Experiment with Omni Reference — Create a character or object, then reference it across multiple generations with --oref.
6. Master the Editor — Learn Vary Region for fixing specific areas. This is where you go from "good enough" to "exactly what I wanted."

🎲 You're ready! The best way to learn is to generate. Start with simple prompts, iterate relentlessly, and don't be afraid to experiment. Every image teaches you something. Good luck!