Most Midjourney users write basic prompts and get basic results. A few parameters and techniques separate professionals from beginners. Here are 15 that will immediately improve your generations.
Add --style raw to any prompt for more photorealistic, less "AI-looking" results. It's the single most impactful parameter change most users don't know about.
The Fundamentals
1. Use --style raw for Realism
Midjourney's default style applies a layer of artistic interpretation. --style raw removes this, producing images closer to your actual prompt. Essential for product photos, portraits, and anything requiring accuracy.
2. Master Aspect Ratios
--ar 16:9 for YouTube thumbnails, --ar 9:16 for Instagram Stories, --ar 4:5 for Instagram feed, --ar 1:1 for profile pictures. Always specify — the default square rarely fits your actual use case.
3. Control Quality with --q
--q 2 doubles the generation time but produces noticeably better detail. Use it for final outputs; skip it for ideation. --q .25 generates 4x faster for quick concept testing.
Lighting and Camera Techniques
4. Specify Camera Lenses
Add lens specifications for dramatically better portraits: "shot on 85mm f/1.4 lens" for bokeh, "shot on 24mm wide angle" for landscapes, "100mm macro lens" for detail shots. Midjourney understands photography terminology.
5. Control Lighting Precisely
Instead of "good lighting," specify: "golden hour," "blue hour," "overcast diffused light," "dramatic side lighting," "ring light," or "Rembrandt lighting." Each produces distinctly different moods.
6. Add Film References
"Shot on Kodak Portra 400," "ARRI Alexa," "Fujifilm X100" — camera and film references dramatically affect the mood and grain of your images. Experiment with different film stocks for different vibes.
Style Control
7. Reference Specific Artists (for Style)
Add "in the style of [artist]" for targeted aesthetics. Note: this references style, not reproduces copyrighted work. "In the style of Monet" gives impressionist brushwork; "in the style of Ansel Adams" gives dramatic black and white photography.
8. Use --stylize for Artistic Intensity
--stylize 0 follows your prompt literally with minimal artistic interpretation. --stylize 1000 goes full artistic. Default is 100. Try 50 for more accurate product shots, 500 for more artistic portraits.
9. Combine Multiple References
Mix styles: "cinematic photography meets editorial fashion, golden hour, 85mm lens, desaturated film grain, by Annie Leibovitz." Layering multiple references creates unique aesthetics.
Advanced Techniques
10. Use Negative Prompts Wisely
Add --no [element] to exclude things: --no text, watermark, signature, blurry, low quality should be in almost every prompt. --no people for pure product shots.
11. Seed for Consistency
Add --seed [number] to reproduce similar compositions. Generate an image you like, get its seed from the envelope emoji reaction, use the same seed with modified prompts for consistent characters or environments.
12. Vary Region for Surgical Edits
The Vary (Region) feature lets you repaint specific parts of an image while keeping the rest. Perfect for fixing hands, swapping backgrounds, or changing one element without regenerating everything.
13. Use Image Prompts for Reference
Add an image URL at the start of your prompt: [image URL] [text prompt] --iw 0.5. The --iw parameter controls how much influence the image has (0-2). Use your own product photos as reference for consistent product imagery.
14. Upscale Strategically
Use "Upscale (Subtle)" for images that are already good — it sharpens without changing. Use "Upscale (Creative)" when you want the AI to add more detail and potentially improve the composition.
15. Build a Prompt Library
When you find a combination that works, save it. Your prompt archive is your competitive advantage. Check our Image Generation Prompt Library for battle-tested starting points.
Try this: "Cinematic portrait of a [subject], shot on 85mm f/1.4, golden hour lighting, film grain, Kodak Portra 400, desaturated tones, shallow depth of field --ar 2:3 --style raw --stylize 200 --q 2"
Frequently Asked Questions
Understanding How Midjourney Thinks About Prompts
Midjourney treats prompts more like a weighted combination of concepts and artistic references than as instructions. The order of words matters — earlier concepts get more weight. Adjacent concepts blend together, and specific references to artists, photographers, or film directors strongly steer the aesthetic. "A forest, Thomas Moran, golden hour, oil painting" produces a different result than the same words reordered. The implication for prompting: lead with your most important subject, follow with style and aesthetic descriptors, and end with technical parameters.
The Parameter Toolkit You Should Know
Beyond the basics, these parameters meaningfully improve results: --cref [URL] (Character Reference) maintains a specific character's appearance across multiple generations — essential for any project requiring consistent characters. --tile generates seamlessly repeating patterns for textures and backgrounds. --iw [0-3] controls how much an image reference influences the output versus the text prompt. --chaos [0-100] controls variation between the four initial options — high chaos for exploring unexpected directions, low chaos for consistent execution. --repeat [n] automatically runs a prompt multiple times for generating large batches. Learning these 5-6 parameters separates casual users from power users and opens up creative possibilities not obvious from the basic interface.
Style Reference Library: Building Your Visual System
The most underutilized Midjourney workflow for professional work is building a style reference library. Generate 5-10 images with a consistent aesthetic you love, identify the 2-3 that best capture your target look, save those URLs, and use them consistently as --sref references across projects. This creates visual coherence across a campaign or portfolio that exhaustive prompt repetition can't match. For brand work: create your "style bible" by generating reference images for your brand's visual identity, then use these as sref references for every subsequent brand generation. The consistency is significantly better than trying to describe the same aesthetic in text each time.
Fixing the Most Common Quality Problems
Extra limbs and distorted anatomy: Add "--no extra limbs, deformed hands, anatomical errors" and use the V6 model which has significantly better anatomy. Generic, over-polished results: Add "--style raw" to reduce Midjourney's tendency to over-beautify outputs. This produces more photorealistic, gritty results. Combine --style raw with specific photographer references (Annie Leibovitz, Steve McCurry) for results with genuine photographic character rather than AI-typical glossiness. Inconsistent lighting: Be explicit and specific: "single light source from upper left, hard shadows, high contrast" rather than "dramatic lighting." Cinematography references work well: "Roger Deakins cinematography, Blade Runner 2049 color grading" encodes a complex lighting aesthetic compactly.
Using Midjourney for Professional Client Work
The workflow that works best for professional client work: generate 20-30 variations in an exploration phase using varied prompts and high --chaos values to find unexpected directions, present the top 5-8 as direction options to the client, then use the approved direction with low --chaos and --sref references to produce consistent final deliverables. This two-phase approach (exploration then execution) matches how traditional creative briefing works and produces better client outcomes than either over-constraining the initial exploration or delivering whatever the AI generates without curation.