All guides
BeginnerSelf-pacedFree GuideSection 33/50

Midjourney — Advanced techniques

Explore generative AI through prompts, assistants, multimedia and response checking. Work through sections and exercises at your own pace.

💡

Free content, available without an account. Tools and models change: use the current model guide to check their versions. Models and sources

Section 10.5.2 : Midjourney — Advanced techniques

🎯 Learning objective

Leverage Midjourney to create high aesthetic quality visuals through advanced parameters, style references, and visual prompting techniques. You'll master the complete professional workflow, from creative exploration to production-ready final files.


Midjourney: the king of aesthetics

Midjourney is the image generation model most used by professional creatives. Its strength: exceptional aesthetic quality "out of the box" and granular control via advanced parameters. Midjourney images are immediately recognizable for their visual quality — it's often the choice of art directors, designers, and photographers.

Midjourney works via Discord (or the website midjourney.com since 2024). Pricing: from $10/month (Basic) → $30/month (Standard, recommended) → $60/month (Pro).

Basic syntax and complete parameters

The basic command: /imagine [your prompt in English] --parameters

/imagine a modern workspace with warm lighting, minimalist design --ar 16:9 --v 7 --style raw --s 250

All important parameters:

ParameterMeaningValuesWhen to use
--arAspect ratio1:1, 16:9, 9:16, 3:2, 4:5Always specify it (otherwise defaults to 1:1)
--vModel version7 (latest, April 2025), 6.1, 5.27 for best quality
--style rawLess AI embellishmentOn/offWhen the result is "too pretty" and lacks naturalness
--s (stylization)Artistic intensity0 to 1000 (default: 100)0 = faithful to prompt, 1000 = very aesthetic
--c (chaos)Variety/surprise0 to 100 (default: 0)Exploration phase: 30-50. Refinement phase: 0-10
--q (quality)Render quality0.25, 0.5, 1 (default), 22 for final render, 0.5 for quick tests
--noExclude elementsWords separated by commas--no text, watermark, people, border
--seedRandom seedNumberSame seed + same prompt = same result (reproducibility)
--tileSeamless/tileable imageOn/offRepetitive patterns, textured backgrounds
--w / --hWidth/height (legacy)PixelsPrefer --ar

Advanced prompting techniques

1. Style References (--sref) — The most powerful:

Have an image whose style you love? Use it as a reference:

/imagine futuristic city skyline at sunset --sref [image URL] --sv 100
  • →--sv 0: subtle influence from the reference style
  • →--sv 100: faithful copy of the style (default)
  • →You can combine multiple --sref: --sref URL1 --sref URL2

Use case: you created an illustration you love for chapter 1 of your book. You want all subsequent illustrations to have the same style → use the first one as --sref for the entire series.

2. Character References (--cref) — Character consistency:

/imagine [your mascot] in a coffee shop reading a book --cref [character URL] --cw 100
  • →--cw 0: copies only the face
  • →--cw 100: copies the entire character (clothing, posture, style)

Use case: you're creating a series of posts with a recurring character → it stays visually consistent from one image to the next.

3. Multi-prompts with weighting (::):

The :: separates and weights concepts:

/imagine cyberpunk city::2 neon rain::1.5 lonely figure::0.5 --ar 16:9
  • →cyberpunk city::2 = double importance
  • →neon rain::1.5 = 1.5x importance
  • →lonely figure::0.5 = 0.5x importance (subtle, in the background)

Negative weighting (remove a concept):

/imagine beautiful garden::2 flowers::1 insects::-0.5

The -0.5 reduces the probability of seeing insects.

4. Permutation prompts (bracketed options):

Test multiple variants in a single prompt:

/imagine {a cat, a dog, a robot} sitting on a {red, blue, green} chair

Generates 9 images: all combinations of [subject] × [color].

5. Prompt stacking for hybrid styles:

/imagine [subject], in the style of [artist/movement], 
blended with [other style], [specific color palette],
[type of lighting], [mood], award-winning, highly detailed

Concrete example:

/imagine ethereal portrait of a jazz musician, in the style of 
renaissance oil painting, blended with cyberpunk neon aesthetics, 
palette of deep blues and warm golds, dramatic chiaroscuro lighting, 
atmospheric and mysterious, masterpiece quality --ar 3:4 --v 7 --s 400

Professional Midjourney workflow in 6 steps

StepObjectiveParametersNumber of images
1. ExplorationFind a creative direction--c 50 --s 250 --q 0.516-20 images (4 batches)
2. DirectionChoose the preferred style--c 20 --s 1508 images
3. RefinementFine-tune the rendering--c 0 --s 100 --style raw8 images
4. VariationsSubtle variations of the favoriteV1-V4 buttons4 images
5. UpscaleHigh resolutionU1-U4 button → "Upscale (Subtle)"1 image
6. Post-productionCleanup and adjustmentsExport + Photoshop/Canva1 final file

Total time: 15-30 minutes for a professional visual (vs 2-4h with a photographer + traditional retouching).

Common mistakes and how to avoid them

MistakeProblemSolution
Prompt in FrenchLower quality resultsAlways prompt in English (Midjourney is trained on English data)
Prompt too long (> 100 words)Midjourney "drowns" elementsPrioritize: the first 20 words have the most impact
No --ar specifiedSquare image by defaultAlways specify the desired format
--s too high"Too AI" look, unrealisticUse --s 50-150 for a natural rendering
Ignoring --style rawExcessive embellishmentEnable --style raw for realistic photos

DALL-E 3 vs Midjourney v7 — Detailed comparison

DALL-E for quick daily use, Midjourney for demanding creative projects
CriterionDALL-E 3Midjourney v7
Aesthetic qualityGoodExceptional
Prompt fidelityExcellentExcellent (v7)
Text in imageGood (not perfect)Improved (v7, not perfect)
Ease of useVery easy (natural chat)Moderate (Discord + parameters)
IterationExcellent (conversational)Good (V1-V4, --sref)
Artistic controlLimited (no parameters)Very advanced (--sref, ::, --s, --cref)
Series consistencyDifficultExcellent (--sref + --cref + --seed)
PriceIncluded in ChatGPT Plus ($20)$10-60/month separate

← Scroll horizontally →

Quick Check

What does the --sref parameter do in Midjourney?

Quick Check

Which Midjourney parameter should you use during the creative exploration phase?

Your progress

Progress is self-reported and stored in your browser. The final completion record is not a qualification or an independent assessment.

View progress and completion record

Learn AI, from prompts to agents

15 free guidesQuizzes, diagrams and workshopsNo account needed

One practical AI workflow in your inbox

Methods, resources and new exercises. Free, unsubscribe at any time.