Image Generation / Midjourney
Determine –c value based on desired output diversity — calibrates variation scientifically, not randomly.
Why This Prompt Exists
–chaos (–c) controls how much variation you get between generations. Most users leave it at 0 (identical outputs) or crank it to 100 (unusable randomness) — missing the productive middle range.
You get:
- chaos = 0: every generation looks the same (no exploration)
- chaos = 100: outputs are unrecognizably different (no consistency)
- no understanding of what different chaos values actually produce
- wasted generations because chaos is set wrong for the task
- inability to explore variations while keeping brand identity
But chaos values have predictable effects:
- 0-10: Very similar outputs, minor variations in details
- 11-30: Noticeable variations in composition and framing
- 31-60: Significant variations, different interpretations of the prompt
- 61-80: High variation, may change subject interpretation
- 81-100: Extreme variation, unpredictable results
Without calibration, chaos is a guessing game.
This prompt recommends chaos values by intended variation level.
The Prompt
Assume the role of a Midjourney variation scientist who calibrates chaos parameters. Your task is to recommend --c values based on desired output diversity. Generate: 1. VARIATION GOAL CLASSIFICATION - Desired consistency: [IDENTICAL / SIMILAR / MODERATE VARIATION / HIGH VARIATION / EXPLORATION] - Use case type: [BRAND ASSET / CONCEPT EXPLORATION / SERIES / ONE-OFF] 2. CHAOS VALUE RECOMMENDATION TABLE | Variation Goal | Recommended --c | Expected Output | Best For | |----------------|-----------------|-----------------|----------| | Identical | 0 | Same image, same seed | Reproducible assets | | Very similar | 1-10 | Minor detail changes | Logo variants, small adjustments | | Noticeable variation | 11-30 | Composition changes, different framing | Multiple options for same brief | | Significant variation | 31-60 | Different interpretations | Concept exploration | | High variation | 61-80 | May change subject interpretation | Brainstorming, early ideation | | Extreme exploration | 81-100 | Unpredictable, experimental | Creative inspiration only | 3. USE CASE MAPPING | Use Case | Recommended --c | Rationale | |----------|-----------------|-----------| | Brand logo exploration | 30-50 | Need variety but maintain recognizability | | Character concept art | 40-70 | Explore different looks for same character | | Product photography | 0-10 | Consistent product representation | | Social media content | 10-30 | Fresh looks, same style | | Print ad series | 5-20 | Consistent but not identical | | Mood board creation | 50-80 | High variation for inspiration | 4. CHAOS COMBINATION WITH OTHER PARAMETERS - --c + --seed: fixed seed + chaos = variations on same theme - --c + high --stylize: unpredictable (use lower --c) - --c + --no: chaos may override negative prompts at high values 5. CHAOS CALIBRATION TEST PROTOCOL - Step 1: Run 4 generations at --c 0 (baseline) - Step 2: Run 4 generations at --c 25 (moderate variation) - Step 3: Run 4 generations at --c 50 (high variation) - Step 4: Select best value based on output 6. COMMON MISTAKES TO AVOID - Using --c 0 for exploration (wasted generations, no variety) - Using --c 100 for brand assets (inconsistent, unusable) - Expecting chaos to create variation within one generation (it affects multiple runs) - Combining high chaos with high stylize (too unpredictable) INPUTS: Desired consistency level: [IDENTICAL / SIMILAR / MODERATE / HIGH / EXTREME] Use case: [E.G., "Brand logo exploration", "Product photos", "Concept art"] Number of variations needed: [E.G., "10-20 options to choose from"] Subject type: [E.G., "Abstract", "Realistic product", "Character", "Landscape"] RULES: - --c controls variation across multiple generations, not within one image - Lower chaos (1-10) for brand consistency where identity matters - Medium chaos (30-50) for concept exploration where variety is valuable - High chaos (70-100) for brainstorming only — not final assets - Test chaos at 25, 50, and 75 to understand your specific prompt's sensitivity - Some prompts are more chaos-sensitive than others (abstract prompts vary more)
How To Use It
- –c controls variation across multiple generations, not within a single image.
- Lower chaos (1-10) for brand consistency where identity matters.
- Medium chaos (30-50) for concept exploration where variety is valuable.
- High chaos (70-100) for brainstorming only — not for final assets.
- Test chaos at 25, 50, and 75 to understand your specific prompt’s sensitivity.
- Some prompts are more chaos-sensitive than others (abstract prompts vary more).
Example Input
Desired consistency level:
“MODERATE — want different options but same character”
Use case:
“Character concept art for a game protagonist”
Number of variations needed:
“20-30 options to choose from”
Subject type:
“Character — fantasy warrior”
Why It Works
Most users treat chaos as a binary choice — off or maximum — missing the graduated scale that makes it useful.
This framework improves outcomes by forcing:
- variation goal classification (identical to extreme exploration)
- chaos value mapping (specific ranges for specific outcomes)
- use case alignment (brand assets vs. concept exploration)
- parameter interaction awareness (chaos + stylize, chaos + no)
- calibration test protocol (finding the right value for your prompt)
Failure modes this prevents:
- Chaos = 0 for exploration (all outputs identical, no variety)
- Chaos = 100 for brand assets (outputs unrecognizable, unusable)
- Expecting chaos to vary within one generation (it affects multiple runs)
- Combining high chaos with high stylize (too unpredictable)
This improves on: Random chaos guessing. Calibrated variation produces useful diversity without unusable outputs.
Related to: MJ-01 (Parameters) for chaos syntax; MJ-06 (Remix) for variation patterns.
Build Better AI Systems
Subscribe for advanced prompt engineering, AI coding tools, debugging frameworks, and practical strategies for developers and engineers.