Midjourney is one of the most powerful neural networks for image generation. It creates photorealistic pictures, illustrations, concept art, and designs from text descriptions. This guide breaks down the entire process from registration to advanced techniques — step by step.
Step 1. Registration and Subscription
Registering on Discord
Midjourney works through Discord — a messenger popular among gamers and developers.
- Go to discord.com and create an account if you don't have one.
- Confirm your email and set up your profile.
- Download the Discord desktop app or use the web version.
Connecting to Midjourney
- Go to midjourney.com.
- Click "Join the Beta" or "Sign In".
- Authorize using your Discord account.
- You will land on the Midjourney server in Discord.
Pricing Plans
| Plan | Price/Month | Fast Generations | Features |
|---|---|---|---|
| Basic | $10 | ~200 images | Basic access |
| Standard | $30 | ~900 images | Unlimited Relax mode |
| Pro | $60 | ~1800 images | Stealth mode, more concurrent tasks |
| Mega | $120 | ~3600 images | Maximum speed and concurrency |
Tip for beginners: Start with the Basic plan for $10 — it's enough to understand the tool and decide if you need more volume.
Step 2. First Generation
Where to Generate
There are two ways to work with Midjourney:
Via the Discord server:
- Go to one of the
#newbieschannels on the Midjourney server. - Your generations will be visible to everyone (except Pro/Mega with Stealth mode).
Via direct messages with the bot (recommended):
- Find the Midjourney bot in the server's member list.
- Click on it and select "Send Message".
- Generations will be private and won't get lost in the stream of others' work.
The /imagine Command
The main command for generating images:
/imagine prompt: your image description
Type /imagine in the chat, Discord will show a hint — click on it and enter your description in the prompt field.
Example of a first prompt:
/imagine prompt: a cozy coffee shop interior, warm lighting, rain outside the window, watercolor style
Within 30–60 seconds, the bot will return a grid of 4 images.
Buttons Under the Result
After generation, buttons will appear under the image:
- U1, U2, U3, U4 — Upscale (enlarge the selected option to full resolution).
- V1, V2, V3, V4 — Variations (create variations based on the selected one).
- 🔄 — Regenerate all 4 options with the same prompt.
Numbering of options: 1 — top left, 2 — top right, 3 — bottom left, 4 — bottom right.
Step 3. Writing Prompts
Prompt quality is 90% of the result. Let's break down how to compose effective descriptions.
Prompt Structure
A good prompt consists of several elements:
[Subject], [details and actions], [environment/background], [style], [lighting], [mood]
Example:
/imagine prompt: a samurai warrior standing on a cliff, cherry blossoms falling, Mount Fuji in the background, dramatic sunset lighting, cinematic composition, highly detailed, 8k
What Works Well
- Specific descriptions — "a ginger cat on a windowsill" is better than "a cat".
- Specifying style — oil painting, watercolor, digital art, photograph, anime, pixel art.
- Describing lighting — golden hour, studio lighting, neon glow, soft ambient light.
- Specifying camera — wide angle, close-up, bird's eye view, macro shot.
- Technical terms — 8k, highly detailed, cinematic, photorealistic.
What to Avoid
- Overly long prompts — Midjourney works better with concise and precise descriptions.
- Negative formulations — "without people" might not work. Better describe what you want to see.
- Abstract descriptions — "something beautiful and inspiring" will give unpredictable results.
- Listing many objects — the more elements, the worse the neural network processes each one.
Step 4. Generation Parameters
Parameters are added at the end of the prompt and start with --.
Aspect Ratio (--ar)
Sets the image proportions:
--ar 16:9 — widescreen (landscapes, covers)
--ar 9:16 — vertical (stories, Pinterest)
--ar 1:1 — square (Instagram, avatars)
--ar 3:2 — classic photo
--ar 2:3 — portrait photo
Example:
/imagine prompt: mountain landscape at sunset, dramatic clouds --ar 16:9
Model Version (--v)
Selects the Midjourney algorithm version:
--v 6.1 — current version (recommended)
--v 6 — previous stable version
--niji 6 — model optimized for anime and illustrations
Stylization (--stylize or --s)
Controls how "artistically" Midjourney interprets the prompt:
--s 0 — maximum adherence to the prompt
--s 100 — default value, balanced
--s 250 — moderate artistic freedom
--s 750 — strong stylization
--s 1000 — maximum artistic interpretation
Quality (--q)
Determines generation time and detail:
--q .25 — fast generation, fewer details (for experiments)
--q .5 — medium quality
--q 1 — standard quality (default)
--q 2 — enhanced detail (uses more GPU minutes)
Chaos (--chaos or --c)
Controls result diversity:
--c 0 — results are similar to each other (default)
--c 50 — moderate diversity
--c 100 — maximally different options
Prompt Weights
Use :: to indicate the importance of prompt parts:
/imagine prompt: cat::2 wearing a hat::1 in a garden::.5
The higher the number after ::, the more attention Midjourney pays to that part. Negative values work as exclusions.
Step 5. Advanced Techniques
Remix Mode
Remix allows you to change the prompt when creating variations. Enable it via /settings → Remix Mode.
How to use:
- Generate an image.
- Press V1–V4 for variations.
- In the window that appears, change the prompt.
- Get variations with the new description but preserved composition.
Example: generated a summer landscape, via Remix change "summer" to "winter" — get the same angle but in winter.
Vary (Region)
Allows redrawing part of an image. After Upscale, click "Vary (Region)", select an area, and enter a new prompt — Midjourney will redraw only the selected part.
Pan and Zoom Out
Pan — extends the image beyond its borders. Click the direction arrows after Upscale, and Midjourney will draw additional content. Zoom Out — zooms out the camera and draws the surroundings (1.5x and 2x buttons available).
Reference Images
Upload an image as a reference by inserting its URL before the description. The --iw parameter (from 0 to 2) controls the degree of the reference's influence on the result.
Step 6. Practical Tips
Iterative Approach
Don't try to get the perfect result on the first try:
- Draft — short prompt,
--c 50for diversity. - Refinement — choose the best option, add details.
- Variations — use V1–V4 for refinement.
- Final — Upscale the best result.
Useful Styles for Prompts
| Style | Keywords |
|---|---|
| Photorealism | photorealistic, 8k, DSLR, Canon EOS R5, 85mm lens |
| Cinematic | cinematic, film grain, dramatic lighting, anamorphic |
| Illustration | digital illustration, concept art, artstation |
| Painting | oil painting, watercolor, acrylic, impasto |
| Minimalism | minimalist, clean, simple, flat design |
| Retro | vintage, retro, 1980s, VHS aesthetic, synthwave |
| Anime | anime style, Studio Ghibli, manga, cel shading |
Saving and Organizing
All generations are saved on midjourney.com/app. Use /describe to analyze existing images and save successful prompts in notes for reuse.
Common Beginner Mistakes
- Overly complex prompts. Start with simple descriptions and gradually add details.
- Ignoring parameters.
--ar,--s, and--qdrastically affect the result — experiment. - One attempt = final. Professionals generate dozens of variations for the desired result.
- Prompts only in Russian. Midjourney understands English better — use a translator.
- Copying others' prompts without understanding. Understand each element before copying.
- Not using Remix. Remix Mode is a powerful tool for iterative refinement.
Conclusion
Midjourney is a tool that unfolds with practice. Start with simple prompts, experiment with parameters, study styles and techniques. After 20–30 generations, you'll begin to understand how the neural network "thinks," and your results will become significantly better. The main rule — don't be afraid to experiment. Each generation costs pennies, but the experience you gain is priceless. Save successful prompts, analyze others' work, and constantly try new approaches — that's how truly impressive images are born.