How to Use Midjourney: The Complete Guide to AI Image Generation
Create stunning AI artwork with text prompts — from photorealistic images to painterly illustrations and abstract designs.

Industry-leading AI image generation
Text-to-image from descriptive prompts
Web editor with inpainting and outpainting
Character and style reference consistency
Aspect ratio and stylize controls
Commercial usage on all paid plans
Permutation prompts for batch generation
Describe feature for reverse-engineering prompts
Why This Matters
What sets Midjourney apart is its aesthetic quality. While other generators focus on photorealism alone, Midjourney excels at artistic styles, cinematic lighting, and compositions that feel intentionally crafted rather than randomly assembled. It's particularly strong at concept art, product mockups, editorial illustrations, and stylised portraits.
The platform runs entirely through Discord (with a newer web interface in alpha), which means there's a learning curve around commands and parameters. But once you understand the basics, you can iterate on ideas faster than any traditional design workflow.
Open Midjourney →
How to Do It
Create a Midjourney account
Start with /imagine
/imagine followed by your prompt. Start simple: describe what you want to see, the style, and the mood. For example: /imagine a mountain lake at sunrise, watercolour painting style, serene atmosphere.Add parameters for control
-
--ar 16:9 sets the aspect ratio-
--v 6.1 selects the model version-
--style raw reduces default stylisation-
--no [element] excludes unwanted elementsIterate with variations
Refine with advanced techniques
--iw to control how much influence the reference image has. Combine with --seed values to maintain consistency across related images.Organise and download
What This Actually Looks Like
The Prompt
A cozy Japanese ramen shop at night, rain on the windows, warm interior glow, Studio Ghibli style, detailed, atmospheric, soft lighting, steaming bowls on the counter --ar 16:9 --v 6.1 --niji 6
Example output — your results will vary based on your inputs
After selecting V2 for a variation with the best composition, then U2 to upscale, the final image is a 1024x576 illustration suitable for a blog header, social media post, or presentation slide.
How to Edit This
--niji 6 if you want Midjourney's standard model instead of the anime-focused one. Add --chaos 30 to get more varied and unexpected results across the grid.Prompts to Try
Photorealistic Product Shot
A premium wireless headphone on a marble surface, soft studio lighting, shallow depth of field, commercial photography style, 8k resolution --ar 16:9 --v 6.1 --style raw
What to expect: A clean, professional product image suitable for e-commerce or marketing materials. The --style raw flag reduces Midjourney's default stylisation for a more natural look.
Illustrated Character Concept
A friendly robot librarian organizing glowing books on floating shelves, Studio Ghibli inspired, warm colour palette, detailed background, whimsical atmosphere --ar 3:4 --v 6.1 --niji 6
What to expect: An anime/illustration-style character with rich detail. The --niji flag activates Midjourney's anime model, ideal for illustrated content and character design.
Architecture Visualisation
Modern eco-friendly house with living green walls, floor-to-ceiling windows overlooking a forest, golden hour sunlight, architectural photography, shot on Hasselblad --ar 16:9 --v 6.1
What to expect: A photorealistic architectural render with natural lighting. Mentioning specific camera brands helps Midjourney emulate professional photography characteristics.
Common Mistakes
Vague prompts with no style direction
Ignoring aspect ratio
--ar 16:9 for presentations and headers, --ar 9:16 for mobile and stories, --ar 3:4 for portraits. Always set your aspect ratio intentionally.Not using negative prompts
--no to exclude them. For example, --no text, watermark, blurry removes common artefacts that degrade quality.Skipping the variation and upscale workflow
Overloading prompts with contradictory styles
Tools That Work for This
The primary interface — type /imagine in any Midjourney Discord channel to generate images. Supports parameters like --ar, --v, --style, --no, and --chaos.
A newer browser-based interface at midjourney.com/imagine for generating and managing images without Discord. Currently in alpha with expanding features.
Built-in /describe command that analyses an uploaded image and suggests prompts to recreate it — excellent for reverse-engineering styles you admire.
Midjourney's specialised anime and illustration model, activated with --niji 6. Produces distinctly different results optimised for illustrated and character-focused content.
Getting Started: Setup, Subscription, and Your First Image
Once you're in, the workflow is straightforward: type
/imagine followed by your description in any Midjourney bot channel. The bot generates a grid of 4 images based on your prompt within about 60 seconds.For your first prompt, try something descriptive:
/imagine a cozy bookshop interior at golden hour, warm lighting, wooden shelves filled with books, watercolour painting styleKey controls to learn immediately:
- V1-V4 buttons create variations of a specific image from the grid
- U1-U4 buttons upscale an image to full resolution
-
--ar 16:9 sets the aspect ratio (crucial for real-world use)-
--v 6.1 selects the model version-
--style raw reduces Midjourney's default aesthetic stylisationThe best results come from iteration. Don't expect perfection from the first grid — use variations to refine towards what you're envisioning, adjusting your prompt each round.
Prompt Engineering and Essential Parameters
Image Prompts let you upload a reference image (paste the URL before your text) and Midjourney will incorporate its style, composition, or colours. Use
--iw (image weight) to control how strongly the reference influences the output — from subtle inspiration to close replication.Remix Mode allows you to change your prompt between variations. Enable it with
/settings and toggle Remix on. Now when you click V1-V4, you can modify the prompt while keeping the composition — perfect for trying different styles on the same scene.Multi-Prompts use
:: to separate concepts with different weights. For example, hot dog::2 food art::1 emphasises the food aspect over the art direction. This gives you precise control over how Midjourney balances competing elements.The /describe Command is an underrated learning tool — upload any image and Midjourney will suggest 4 prompts that could recreate it. It's the fastest way to learn what language and descriptors produce specific visual effects.
Consistent Characters using
--cref (character reference) lets you maintain the same character across multiple generations — essential for storytelling, branding, and series work.Advanced Techniques and Professional Workflows
Structure your prompts in layers. Follow this formula: Subject + Setting + Style + Lighting + Technical specs. For example: 'a samurai warrior (subject) standing on a misty mountain ridge (setting), ukiyo-e woodblock print style (style), dramatic backlit dawn (lighting) --ar 16:9 --v 6.1 (technical)'.
Use specific art and photography references. 'Shot on Hasselblad', 'Kodak Portra 400 film', 'in the style of Studio Ghibli', or 'Bauhaus design' all produce dramatically different results. Midjourney understands hundreds of artistic references.
Control mood with lighting descriptors. 'Golden hour', 'moody chiaroscuro', 'neon noir', 'soft diffused studio lighting', and 'dramatic rim lighting' each transform the same scene entirely. Lighting is often more important than the subject itself.
Use negative prompts strategically. Add
--no text, watermark, blurry, distorted to exclude common artefacts. For photorealistic work, --no illustration, cartoon, painting helps keep the output grounded.Build a personal prompt library. When you discover prompt formulas that consistently produce great results, save them. The compound value of a well-organised prompt library far exceeds any single generation.
