Midjourney from Zero
Jump to section
Why Midjourney is the best starting point
Midjourney is an AI image generation tool that specializes in artistic, beautifully stylized visuals. Unlike other tools, Midjourney has a strong aesthetic opinion — even simple prompts produce stunning results. It is the perfect gateway into AI-generated imagery for anyone in a creative field.
You can use Midjourney in two ways: through Discord (the original method, using the /imagine command) or through the web interface at midjourney.com. The web interface is now the recommended approach — it offers a gallery, generation history, and a more comfortable editing experience. Both require a paid subscription starting around $10/month.
If you are new to the platform, skip Discord entirely and go straight to midjourney.com. Sign in, click 'Create', type your prompt, and hit Enter. You will have your first image in under a minute.
Your first image
Type a description of what you want to see, and Midjourney generates four variations in a 2x2 grid. You do not need a perfect prompt — start simple and iterate. Here is your very first one:
/imagine a cozy bookstore interior with warm lamp light and tall wooden shelvesFrom the grid, you can Upscale (U1-U4) to get a high-resolution version of one image, or create Variations (V1-V4) to generate similar alternatives. Pick your favorite and upscale it to see the full detail.
The anatomy of a great prompt
A well-structured Midjourney prompt covers five dimensions: subject (what), style (how it looks), medium (the technique), lighting (the light source and quality), and mood (the emotional feel). You do not need all five every time, but the more you specify, the more predictable your results.
/imagine a Japanese garden in autumn, ukiyo-e woodblock print style, soft diffused light, contemplative atmosphere --ar 16:9/imagine portrait of a jazz musician playing saxophone, cinematic photography, dramatic side lighting, smoky club atmosphere --ar 3:4- Subject: What is in the image (person, landscape, object, scene)
- Style: Visual approach (photorealistic, watercolor, comic book, cinematic)
- Medium: The technique (oil on canvas, digital painting, 35mm film photography)
- Lighting: Light quality (golden hour, dramatic shadows, soft studio light)
- Mood: Emotional tone (melancholic, energetic, serene, mysterious)
Essential parameters
Parameters go at the end of your prompt after a double dash (--). They control the technical aspects of generation without changing the creative content. Here are the ones you will use most often:
- --ar 16:9 — aspect ratio (landscape). Other common ratios: 1:1, 3:4, 9:16
- --style raw — disables Midjourney's aesthetic enhancements, giving you output closer to what you described
- --stylize 50 — controls how much Midjourney adds its own artistic flair (0-1000, default 100). Lower = faithful to prompt, higher = more creative
- --chaos 30 — controls variety between the four variations (0-100). Higher = more diverse results
- --no text, watermark — negative prompt specifying what you do NOT want in the image
/imagine clean minimalist product photography of a ceramic coffee mug, white background, soft shadows --ar 1:1 --style raw --stylize 20 --no text, logo, brandingThe --style raw parameter is essential for commercial work. Without it, Midjourney adds its signature look — beautiful but not always appropriate for brand assets. With --style raw, the output stays closer to your actual description.
Where Midjourney shines and where it struggles
Midjourney excels at artistic illustrations, concept art, fantasy and sci-fi scenes, fashion photography, architectural visualization, and any scenario where aesthetic beauty matters more than photographic accuracy. Even a simple three-word prompt can produce gallery-worthy results.
Midjourney struggles with: text in images (letters are often garbled), precise anatomy (hands, fingers), specific real faces, and exact object counts. For text rendering, use DALL-E. For photorealistic portraits, try Flux. Know each tool's strengths.
Choose a single subject (for example, 'an abandoned lighthouse on a cliff') and generate 5 images using different artistic styles. Start with the prompts below and customize them.
Hint
Try: watercolor, oil painting, Studio Ghibli anime, cyberpunk neon, vintage photograph. Notice how the style completely transforms the mood and story of the same scene.
/imagine an abandoned lighthouse on a cliff, watercolor style, misty morning light --ar 16:9
/imagine an abandoned lighthouse on a cliff, oil painting, stormy sky, dramatic waves --ar 16:9
/imagine an abandoned lighthouse on a cliff, Studio Ghibli anime style, lush green grass, magical --ar 16:9
/imagine an abandoned lighthouse on a cliff, cyberpunk style, holographic beams, neon fog, night --ar 16:9
/imagine an abandoned lighthouse on a cliff, vintage 1960s photograph, black and white, film grain --ar 16:9Take your favorite prompt from the previous exercise and generate it three times: once with --stylize 20, once with --stylize 100 (default), and once with --stylize 750. Compare the results side by side.
Hint
Low stylize (20) stays faithful to your description. High stylize (750) gives Midjourney creative freedom to add dramatic effects you did not ask for. Notice which version you actually prefer.
Generate a single visual in three formats: --ar 1:1 (Instagram post), --ar 9:16 (Instagram Story or Reels), and --ar 16:9 (YouTube thumbnail). Observe how Midjourney adapts the composition to fit each ratio.
Hint
Midjourney automatically adjusts composition — vertical ratios produce taller subjects, wide ratios give more environmental context. You may need to tweak your prompt to get the best result in each format.
- Midjourney excels at artistic, stylized visuals — beautiful results even from simple prompts.
- Prompt structure: subject + style + medium + lighting + mood = predictable output.
- Key parameters: --ar for aspect ratio, --style raw for prompt fidelity, --stylize for Midjourney's creative input level.
- Use --no to remove unwanted elements from your images.
- Midjourney struggles with text, hands, and precise counts — use other tools for those tasks.