Character Reference (--cref)
Consistent characters across images
Character Reference is revolutionary for storytelling. It allows you to maintain the same character's appearance across multiple generations - perfect for comics, children's books, game assets, and brand mascots.
🛠️ How It Works (Midjourney):
/imagine prompt: [your description] --cref [image URL]
Optional: --cw [0-100] to control character weight
✅ Best Use Cases:
- • Comic book series
- • Children's book illustrations
- • Character design sheets
- • Brand mascots
- • Storyboarding
- • Social media characters
⚙️ Character Weight (--cw):
- • --cw 0 = Face only
- • --cw 50 = Face + some features
- • --cw 100 = Full character (default)
- • Lower = more creative freedom
- • Higher = stricter matching
📋 Example - Same Character, Different Scenes:
"A brave knight in a dark forest, holding a glowing sword, dramatic lighting --cref https://example.com/my-knight.jpg --cw 80 --ar 16:9"
📋 Example - Character Sheet:
"Character sheet, multiple poses, front view, side view, back view, action pose, expressions --cref https://example.com/character.jpg --cw 100"
Style Reference (--sref)
Transfer artistic style without copying content
Style Reference extracts the artistic style, colors, textures, and mood from a reference image and applies it to your new creation. The content remains completely different - only the style transfers.
🛠️ How It Works (Midjourney):
/imagine prompt: [your description] --sref [style image URL]
Optional: --sw [0-1000] to control style weight (default 100)
✅ What Transfers:
- • Color palette
- • Brush strokes / textures
- • Artistic technique
- • Mood and atmosphere
- • Lighting style
- • Overall aesthetic
❌ What Doesn't Transfer:
- • Subject matter
- • Composition
- • Specific objects
- • Characters/people
- • Scene layout
📊 Style Weight Examples:
--sw 50
Subtle influence
--sw 100
Balanced (default)
--sw 500
Strong style match
📋 Example - Apply Van Gogh Style:
"A cozy coffee shop interior with warm lighting --sref https://example.com/starry-night.jpg --sw 200"
📋 Example - Consistent Brand Style:
"Product photography of a smartphone on marble surface --sref https://example.com/brand-style.jpg --sw 150 --ar 1:1"
Image Prompts
Use images as creative input
Image prompts let you include an image as part of your prompt. The AI analyzes the image's composition, colors, subjects, and style, then blends this with your text prompt.
🛠️ How It Works:
Midjourney:
/imagine prompt: [image URL] [your text description]
DALL-E 3:
Upload image → "Create a variation of this..."
Leonardo AI:
Use "Image to Image" feature with strength slider
✅ Great For:
- • Creating variations
- • Mood/color inspiration
- • Composition guides
- • Blending multiple images
- • Concept art exploration
💡 Pro Tips:
- • Use --iw to control image weight
- • Combine multiple images
- • Add text to override elements
- • Lower weight = more creative
📋 Example - Image + Text Blend:
"https://example.com/landscape.jpg a magical castle in the mountains, fantasy art, golden hour lighting --iw 0.5"
📋 Example - Multiple Image Blend:
"https://example.com/photo1.jpg https://example.com/photo2.jpg blend these styles, surreal art"
Face Swap & IP-Adapter
Transfer faces to AI-generated images
Face swap technology lets you place a specific person's face onto AI-generated images. This is powerful for personalized art, but must be used responsibly and ethically.
⚠️ Ethical Considerations:
- • Only use faces of yourself or with explicit permission
- • Never create misleading or harmful content
- • Be aware of deepfake laws in your jurisdiction
- • Respect privacy and consent
🛠️ Tools for Face Swap:
Leonardo AI
- • PhotoReal with Face Reference
- • Upload reference face
- • Adjust similarity strength
- • Works with any prompt
Stable Diffusion
- • IP-Adapter extension
- • ReActor for face swap
- • InstantID for likeness
- • ControlNet + face models
Midjourney
- • Use --cref with --cw 0
- • Focuses on face only
- • Best for stylized results
- • Not photorealistic swap
Dedicated Tools
- • InsightFace
- • FaceSwap
- • Roop (open source)
- • Various online services
📋 Example - Midjourney Face Focus:
"Professional business portrait, wearing suit, office background, confident expression --cref https://example.com/my-face.jpg --cw 0 --stylize 50"
📊 Reference Features by Tool
| Feature | Midjourney | Leonardo AI | Stable Diffusion | DALL-E 3 |
|---|---|---|---|---|
| Character Ref | ✅ --cref | ✅ Character Ref | ✅ IP-Adapter | ❌ No |
| Style Ref | ✅ --sref | ✅ Style Ref | ✅ ControlNet | ⚠️ Limited |
| Image Prompt | ✅ URL | ✅ Image2Image | ✅ img2img | ✅ Variations |
| Face Swap | ⚠️ --cw 0 | ✅ PhotoReal | ✅ ReActor | ❌ No |
| Strength Control | ✅ --cw/--sw | ✅ Slider | ✅ Denoising | ❌ No |
💡 Best Practices for Reference Images
Use high-quality reference images - Clear, well-lit photos produce better results than blurry or dark images.
Match the aspect ratio - Use reference images with similar aspect ratios to your desired output.
Balance reference weight - Start with default weights and adjust up/down based on results.
Combine techniques - Use --cref AND --sref together for character + style consistency.
Build a reference library - Save your best reference images for consistent use across projects.
Ready to Use Reference Images?
Try our prompt generator with built-in reference image support!