How to Use Midjourney: Complete Beginner's Guide 2026

Midjourney generates the highest-quality AI images available in 2026. Getting started is straightforward — but getting consistently great results requires understanding how Midjourney interprets prompts and which parameters to use. This guide walks through everything from account setup to advanced techniques.

Getting Started with Midjourney

Step 1: Create Your Account

Midjourney has transitioned fully to its web interface at midjourney.com. Discord access is still available but no longer required for most users.

1. Visit midjourney.com
2. Sign in with your Google or Discord account
3. Choose a subscription plan (Basic starts at $10/month — no free trial)

The Basic plan provides 200 generations per month, sufficient for learning and casual use. The Standard plan ($30/month) adds unlimited “relaxed” generations — slower but uncapped.

Step 2: Your First Generation

After subscribing, click “Create” or go to the Imagine bar:

1. Type a description of what you want to see
2. Press Enter
3. Wait 30-60 seconds for four image variations

Your first prompt: Try something simple and specific: a fox sitting in a forest, golden hour light, photorealistic, detailed fur

Midjourney generates four variations. You can:

• U1-U4: Upscale any of the four images to higher resolution
• V1-V4: Create four new variations based on that image
• Re-roll: Generate four completely new variations from the same prompt

Understanding Midjourney Prompts

Midjourney prompts are more like painting descriptions than commands. Think about:

• Subject: What is in the image?
• Style: Photorealistic, illustration, oil painting, watercolor, anime?
• Lighting: Golden hour, studio lighting, backlit, dramatic shadows?
• Mood: Serene, dramatic, mysterious, playful?
• Composition: Close-up, wide shot, aerial view, profile?

Prompt Examples

Weak prompt: woman in city

Strong prompt: portrait of a woman in a busy Tokyo street at night, bokeh background, street lights reflecting on wet pavement, cinematic composition, shallow depth of field

The additional detail — location specificity, time of day, lighting conditions, technical photography terms — guides Midjourney toward a specific aesthetic rather than a generic interpretation.

Essential Midjourney Parameters

Parameters are added at the end of your prompt with -- prefix and modify how Midjourney generates the image.

Aspect Ratio --ar

Control the image dimensions:

• --ar 16:9 — Landscape (YouTube thumbnails, desktop wallpapers)
• --ar 9:16 — Portrait (mobile, Instagram Stories, TikTok)
• --ar 1:1 — Square (Instagram posts)
• --ar 4:3 — Traditional photo ratio

Style Strength --stylize or --s

Controls how strongly Midjourney applies its aesthetic preferences:

• --s 0 — Very literal interpretation, minimal stylization
• --s 100 — Default (balanced)
• --s 750 — Highly stylized, strong Midjourney aesthetic

For photorealistic images, lower stylize values often produce more accurate results.

Chaos --chaos or --c

Controls variation between the four generated options:

• --c 0 — Four very similar variations (default)
• --c 50 — Moderate variation
• --c 100 — Maximum variation, very different options

Use higher chaos when exploring a concept. Use lower chaos when you have a result you like and want refinements.

Version --v

Specify the model version:

• --v 6.1 — Current best model (default)
• --v 5.2 — Previous generation, different aesthetic

No --no

Remove elements from generations:

• --no text — Removes text from the image
• --no blur — Prevents blurry results
• --no people — Generates scenes without human subjects

Advanced Technique: Style Reference (--sref)

Style reference is Midjourney’s most powerful consistency tool. Paste an image URL after --sref and Midjourney generates new images matching that image’s visual style.

How to use:

1. Generate or find an image whose style you want to replicate
2. Upload it to a hosting service or use a direct URL
3. Add --sref [URL] to your prompt

Example: If you generate a dark moody portrait you love, use it as a style reference for all subsequent prompts in your project. Your images will maintain consistent lighting, color treatment, and aesthetic.

Use multiple style references: --sref [URL1] [URL2] blends both styles.

Advanced Technique: Character Reference (--cref)

Character reference maintains consistent character appearance across multiple images. Essential for:

• Comic and illustration series
• Game character design
• Brand mascot generation
• Story illustration consistency

How to use:

1. Generate your character’s “base” image
2. Add --cref [URL] to new prompts featuring that character
3. Adjust --cw (character weight, 0-100) to control how strictly appearance is maintained

Midjourney will maintain recognizable features — face shape, hair, general appearance — across different scenes, outfits, and angles.

Getting Consistent Results

Seed Numbers --seed

Every generation has a random seed number. Add --seed [number] to replicate a generation starting point across different prompts.

1. Generate an image you like
2. Click the ⓘ info button to see its seed number
3. Add --seed [that number] to new prompts for similar starting compositions

Remix Mode

Enable Remix Mode in settings to modify a specific variation rather than starting fresh. Select a variation, describe the change, and Midjourney adjusts that image specifically.

Useful for: changing outfit colors, swapping backgrounds, adjusting lighting on an otherwise good image.

Practical Tips for Better Results

Use adjective stacking for atmosphere. Layer atmospheric descriptors: eerie, abandoned, overgrown, misty, gothic gives Midjourney specific instructions for a cohesive mood.

Reference real photographers, artists, and films. Prompts like “in the style of Annie Leibovitz photography” or “cinematic like Blade Runner 2049” give Midjourney precise aesthetic targets.

Use photography terminology for realism. Terms like f/1.8 aperture, 85mm lens, bokeh, RAW, shot on Hasselblad guide Midjourney toward photorealistic output with appropriate technical characteristics.

Negative prompting with --no. List things you explicitly do not want: --no text, watermarks, blur, oversaturation prevents common output issues.

Iterate systematically. Find one element that works, lock it with a style reference or seed, then vary one thing at a time. Random exploration is fun but systematic iteration produces professional results faster.

Common Beginner Mistakes

Vague prompts produce generic results. “beautiful landscape” → a beautiful landscape you could find anywhere. “volcanic landscape at dusk, lava flow meeting ocean, steam clouds, dramatic silhouette, cinematic” → something specific and interesting.

Over-specifying conflicting styles. “photorealistic anime oil painting digital art” — Midjourney struggles to reconcile contradictory style descriptors. Pick a primary style direction.

Ignoring the four variations. New users immediately click U1 on the first image they generate. Look at all four first. V1-V4 variations often contain stronger images than the first thumbnail.

Not using upscaling. U1-U4 upscaling adds significant detail and resolution. Always upscale before downloading and using images.

Prompts to Try Right Now

• minimalist interior design, Scandinavian living room, morning light, warm tones, potted plants, hardwood floors --ar 16:9
• oil painting portrait of an elderly fisherman, weathered face, character study, Rembrandt lighting, masterwork --v 6.1
• isometric illustration of a coffee shop, cozy atmosphere, warm colors, detailed, cute style, birds eye view --ar 1:1
• cinematic photograph of a wolf standing on a mountain ridge, moonrise, dramatic clouds, epic scale --ar 16:9 --s 250

Conclusion

Midjourney’s quality ceiling is the highest of any AI image tool, and reaching it requires learning the prompt language. The techniques in this guide — aspect ratios, style reference, character reference, systematic iteration — are what separate casual Midjourney users from professionals who produce consistently impressive work.

Start with simple prompts and add parameters one at a time as you understand their effects. The Midjourney community on Discord and Reddit shares prompts constantly — studying successful prompts accelerates your learning faster than solo experimentation. For comparison with Leonardo AI as a free alternative, see our Midjourney vs Leonardo AI comparison.