midjourney-replicate-flux

What this skill does

# Midjourney-Style Prompt Generator for FLUX 1.1 Pro

Generate professional, Midjourney-quality image prompts optimized for the `black-forest-labs/flux-1.1-pro` model on Replicate.

## Purpose

Transform basic user image requests into rich, detailed prompts that produce Midjourney-quality results using the FLUX 1.1 Pro model. This skill provides:

- Midjourney aesthetic principles and visual characteristics
- FLUX 1.1 Pro model optimization techniques
- Prompt structure templates and patterns
- Before/after transformation examples
- Direct integration with the Replicate MCP server

## When to Use This Skill

Use this skill when:
- User requests image generation via FLUX or Replicate
- User wants "Midjourney-style" or "cinematic" images
- User provides basic image descriptions that need enhancement
- User asks for prompt optimization or improvement
- User requests professional, editorial, or artistic photography
- User wants to generate high-quality AI images

**Trigger phrases**: "generate an image", "create a photo of", "Midjourney style", "cinematic image", "professional photography", "make me a picture of", "FLUX image generation"

## How to Use This Skill

### Step 1: Understand User Intent

Clarify what the user wants to create. Ask targeted questions if the request is too vague:
- What is the main subject?
- What mood or atmosphere do they want?
- What style or genre (portrait, landscape, product, etc.)?
- Any specific preferences (time of day, colors, composition)?

Keep questions focused and avoid overwhelming the user with too many at once.

### Step 2: Select Appropriate Prompt Pattern

Based on the genre, consult `references/midjourney-style-guide.md` for the relevant prompt pattern:
- **Portrait Photography** - People, characters, headshots
- **Landscape/Environment** - Nature, cityscapes, locations
- **Product Photography** - Objects, commercial shots
- **Architectural** - Buildings, structures, interiors
- **Fashion Editorial** - Models, clothing, styling
- **Conceptual/Artistic** - Abstract, surreal, artistic concepts

Each pattern provides a structured template for building prompts in that genre.

### Step 3: Build Layered Prompt

Construct the prompt following the 5-layer structure from `references/midjourney-style-guide.md`:

**Layer 1: Main Subject & Core Composition** (Required)
```
[Subject] [action/pose/position], [composition placement]
```

**Layer 2: Visual Style & Aesthetic** (Required)
```
[art style/movement], [mood descriptor], [color treatment]
```

**Layer 3: Lighting & Environment** (Highly Recommended)
```
[lighting type], [time of day], [weather/atmosphere]
```

**Layer 4: Technical Details & Quality Enhancers** (Recommended)
```
[camera/lens specs], [depth of field], [quality terms]
```

**Layer 5: Artistic References & Influences** (Optional)
```
[artist name], [art movement], [cultural reference]
```

### Step 4: Apply FLUX Optimizations

Enhance the prompt for FLUX 1.1 Pro by consulting `references/flux-model-optimization.md`:

1. **Add explicit artistic treatment**: FLUX is more literal than Midjourney, so specify "cinematic", "editorial", "artistic photography"

2. **Define color grading clearly**: Don't rely on automatic enhancement; specify exact color treatment

3. **Request composition explicitly**: State framing, perspective, and compositional techniques

4. **Include professional quality markers**: "professional photography", "award-winning", "expert color grading"

5. **Use technical photography anchors**: Camera, lens, aperture specifications ground the image in photographic reality

### Step 5: Verify Quality

Check the enhanced prompt against the quality checklist from `references/midjourney-style-guide.md`:
- [ ] Word count is 40-75 words
- [ ] Main subject is clearly defined
- [ ] Lighting is specified with direction and quality
- [ ] Mood/atmosphere is conveyed
- [ ] Technical photography details are present
- [ ] Color treatment is mentioned
- [ ] Composition guidance is included
- [ ] Quality enhancers are present (2-3 minimum)
- [ ] Prompt flows naturally when read aloud
- [ ] No redundant or contradictory terms

### Step 6: Generate Image via Replicate MCP Server

Use the installed Replicate MCP server to generate the image. Call the `mcp__replicate__create_predictions` tool with these parameters:

```json
{
  "version": "black-forest-labs/flux-1.1-pro",
  "input": {
    "prompt": "[Your enhanced Midjourney-style prompt here]",
    "aspect_ratio": "16:9",
    "output_format": "png"
  }
}
```

**Default configuration for Midjourney-style outputs:**
- **aspect_ratio**: `"16:9"` (widescreen, cinematic format)
- **output_format**: `"png"` (highest quality, lossless)
- **safety_tolerance**: 2 (default, balanced - can increase to 3-4 for artistic content if needed)

**Alternative aspect ratios** (from `references/flux-model-optimization.md`):
- `"1:1"` - Square (social media, portraits)
- `"2:3"` or `"4:5"` - Portrait orientation
- `"3:2"` - Classic photo ratio
- `"21:9"` - Ultra-widescreen (panoramic)

The MCP server will return a prediction object. Monitor the prediction status and provide the user with the output image URL when generation completes.

### Step 7: Offer Variations (When Requested)

If the user requests multiple options or variations, provide 2-3 alternative prompts that emphasize different aspects:
- **Variation 1**: Different lighting (golden hour vs blue hour)
- **Variation 2**: Different composition (wide shot vs close-up)
- **Variation 3**: Different mood (dramatic vs serene)

Each variation should maintain the core subject while exploring different artistic directions.

## Reference Documentation

The skill includes three comprehensive reference files that should be consulted during prompt creation:

### 1. `references/midjourney-style-guide.md`

**When to load**: Always consult when building any Midjourney-style prompt

**Contains**:
- Core Midjourney aesthetic principles
- 5-layer prompt structure template
- Midjourney-quality descriptive vocabulary
- Lighting, composition, mood, color, and technical terms
- Common prompt patterns for each genre
- Artistic movement and style references
- Quality control checklist
- Common mistakes to avoid

**Search patterns** (use for quick access to specific sections):
- "Portrait Photography pattern" - Get portrait template
- "Landscape pattern" - Get landscape template
- "Lighting Terms" - Get lighting vocabulary
- "Color & Tone" - Get color treatment options
- "Quality Enhancers" - Get quality modifiers

### 2. `references/flux-model-optimization.md`

**When to load**: Consult when optimizing prompts for FLUX or troubleshooting results

**Contains**:
- FLUX 1.1 Pro specifications and parameters
- Replicate MCP server usage instructions
- FLUX vs Midjourney adaptation strategies
- Model-specific optimization techniques
- Aspect ratio and output format guide
- Safety tolerance settings
- Prompt engineering strategies for FLUX
- Troubleshooting common issues

**Search patterns**:
- "Aspect Ratios" - Get ratio options
- "FLUX Strengths" - Understand what FLUX does best
- "Adaptation Techniques" - Learn FLUX-specific adjustments
- "Parameter Selection" - Choose optimal parameters
- "Troubleshooting" - Fix common issues

### 3. `references/prompt-examples.md`

**When to load**: When learning prompt transformation or seeking inspiration

**Contains**:
- 16 complete before/after transformation examples
- Examples across all major genres (portrait, landscape, product, etc.)
- Word count verification for each example
- Detailed explanations of why each enhancement works
- Transformation principles summary
- Enhancement formula and process

**Search patterns**:
- "Portrait examples" - See portrait transformations
- "Landscape examples" - See landscape transformations
- "Product Photography examples" - See product transformations
- "Transformation Principles" - Understand the enhancement process

## Practical Workflow

### Example 1: Simple Reque