# AI Studio Assignment - Complete Solution Guide

## 🎯 Assignment Overview
**Position:** Generative AI Workflow Builder / Prompting Specialist  
**Company:** Ritz Media World  
**Objective:** Build a rapid prototype for an internal AI platform to support creative teams

---

## 📋 Part 1: System Design & Planning (45 min)

### Architecture Overview

```
┌─────────────────────────────────────────────────────────┐
│                    AI STUDIO PLATFORM                    │
└─────────────────────────────────────────────────────────┘
                          │
        ┌─────────────────┼─────────────────┐
        │                 │                 │
   ┌────▼────┐       ┌────▼────┐      ┌────▼────┐
   │ Frontend │       │ Backend  │      │ Storage │
   │(Streamlit)│       │(Python)  │      │ (Local) │
   └─────────┘       └─────────┘      └─────────┘
        │                 │                 │
   ┌────▼──────────────────▼─────────────────┐
   │         CORE MODULES                     │
   ├──────────────────────────────────────────┤
   │ 1. Ideation Prompt Board                 │
   │    - Prompt templates                    │
   │    - Version control                     │
   │    - Strategy documentation              │
   │                                          │
   │ 2. Image Generator                       │
   │    - HuggingFace (Flux Schnell)         │
   │    - Stable Diffusion (planned)         │
   │    - Midjourney (future)                │
   │                                          │
   │ 3. Video Generator                       │
   │    - RunwayML Gen-2 (stub)              │
   │    - Stable Video Diffusion (stub)      │
   │                                          │
   │ 4. 3D Asset Creator                      │
   │    - Shap-E (demo)                      │
   │    - Blender integration                 │
   │    - Unity export                        │
   │                                          │
   │ 5. Project Library                       │
   │    - Asset management                    │
   │    - Download/Export                     │
   │    - Organization                        │
   │                                          │
   │ 6. History & Feedback                    │
   │    - Prompt versioning                   │
   │    - Team feedback                       │
   │    - Rating system                       │
   │                                          │
   │ 7. Pipeline & Export                     │
   │    - Blender scripts                     │
   │    - Unity integration                   │
   │    - Workflow automation                 │
   └──────────────────────────────────────────┘
                    │
        ┌───────────┼───────────┐
        │           │           │
   ┌────▼───┐  ┌───▼────┐  ┌───▼────┐
   │External │  │External│  │Export  │
   │  APIs   │  │  Tools │  │Targets │
   └─────────┘  └────────┘  └────────┘
```

### External AI Tools Integration

| Tool | Purpose | Status | Why Chosen |
|------|---------|--------|------------|
| **HuggingFace Flux** | Image generation | ✅ Active | Free tier, fast, high quality |
| **Stable Diffusion** | Image generation | 📋 Planned | Industry standard, customizable |
| **Midjourney** | Premium images | 📋 Future | Best quality, but API limited |
| **RunwayML Gen-2** | Video generation | 🚧 Demo | Leader in AI video |
| **Stable Video Diffusion** | Video generation | 📋 Planned | Open source alternative |
| **Shap-E** | 3D generation | 🚧 Demo | OpenAI's 3D model generator |

### Workflow: Brief to Final Asset

```
1. CREATIVE BRIEF SUBMISSION
   └─> Project created with name, description, target audience
       │
2. IDEATION PHASE
   └─> Prompt Board → Template selection → Strategy notes
       │
3. PROMPT REFINEMENT
   └─> Version control → A/B testing → Team collaboration
       │
4. ASSET GENERATION
   ├─> IMAGE: HuggingFace API → PNG files
   ├─> VIDEO: RunwayML API → MP4 files
   └─> 3D: Shap-E API → FBX/OBJ files
       │
5. REVIEW & FEEDBACK
   └─> Team rates assets → Comments → Iteration requests
       │
6. ITERATION LOOP
   └─> Refine prompts → Re-generate → Compare versions
       │
7. EXPORT & INTEGRATION
   ├─> 3D Assets → Blender (.blend files)
   ├─> 3D Assets → Unity (FBX import)
   ├─> Images → Marketing materials
   └─> Videos → Campaign distribution
       │
8. FINAL DELIVERY
   └─> Campaign launch → Performance tracking
```

### Key Features

1. **Prompt Versioning**: Track every iteration with timestamps
2. **Feedback System**: Team members rate and comment on outputs
3. **Blender Integration**: Direct export of 3D assets to .blend format
4. **Unity Support**: FBX export for interactive experiences
5. **Asset Curation**: Library with search, filter, download

---

## 🎨 Part 2: Prompt Engineering Exercise (45 min)

### Luxury Real Estate Marketing Prompts

#### Image Prompt 1: Exterior Shot
**Prompt:**
```
Ultra-modern luxury villa at golden hour, infinity pool with crystal-clear reflections, 
sleek glass facade with floor-to-ceiling windows, lush tropical landscaping, 
ocean view in background, palm trees swaying, warm ambient lighting, 
photorealistic, architectural photography, 8K resolution, professional composition
```

**Strategy:**
- **Time of Day**: Golden hour creates premium aesthetic
- **Technical Details**: Specifies materials (glass, water reflections)
- **Composition**: Clear subject hierarchy (villa primary, ocean secondary)
- **Quality Markers**: "photorealistic", "8K", "professional"
- **Use Case**: Property listing hero image, marketing brochures
- **Integration**: Can be used as reference for 3D environment modeling

---

#### Image Prompt 2: Interior Shot
**Prompt:**
```
Spacious penthouse interior, floor-to-ceiling panoramic windows, 
city skyline at dusk, designer furniture by Italian brands, 
marble flooring with subtle veining, ambient recessed lighting, 
wide-angle perspective, luxury lifestyle aesthetic, high dynamic range, 
interior design magazine quality
```

**Strategy:**
- **Atmosphere**: Lifestyle focus over pure architecture
- **Brand Elements**: "Italian brands" suggests luxury without naming brands
- **Lighting**: "dusk" + "recessed lighting" = warm, inviting
- **Camera**: Wide-angle shows spaciousness
- **Use Case**: Property brochures, virtual staging reference
- **Integration**: Layout can guide 3D interior scene setup

---

#### Video Prompt 1: Aerial Tour
**Prompt:**
```
Cinematic aerial drone footage starting from ocean view, smooth gimbal motion 
flying towards modern beachfront estate, revealing infinity pool and tropical gardens, 
golden hour lighting with warm color grading, 4K resolution, establishing shot 
transitioning to closer detail, professional real estate cinematography, 
10 seconds duration
```

**Strategy:**
- **Camera Movement**: Specific path (ocean → property)
- **Reveal**: Creates anticipation and impact
- **Color Grading**: Warm tones = luxury/comfort
- **Duration**: 10 seconds perfect for social media teasers
- **Use Case**: Instagram Reels, YouTube ads, property website hero
- **Technical**: Smooth motion keywords ensure no jarring movements

---

#### Video Prompt 2: Interior Walkthrough
**Prompt:**
```
Day-to-night time-lapse sequence of luxury condominium lobby, 
marble floors reflecting designer pendant lighting, residents entering and exiting 
in elegant attire, smooth slider camera movement, transitions from natural daylight 
to ambient evening illumination, 15 seconds, atmospheric depth, 
premium hospitality aesthetic
```

**Strategy:**
- **Time-lapse**: Shows building lifecycle and activity
- **Human Element**: "residents" adds life and scale
- **Lighting Transition**: Demonstrates building at different times
- **Movement Type**: Slider = professional, stable
- **Use Case**: Amenity showcase, lifestyle marketing
- **Emotional Goal**: Conveys community and vibrancy

---

#### 3D Asset Prompt: Furniture Piece
**Prompt:**
```
High-polygon 3D model of contemporary outdoor lounge chair, 
teak wood frame with realistic wood grain texture, 
white weather-resistant cushions with subtle fabric weave detail, 
clean geometry suitable for subdivision, PBR materials with proper UV mapping, 
architectural visualization quality, export in FBX and OBJ formats, 
includes diffuse, normal, and roughness maps
```

**Strategy:**
- **Technical Specs**: "high-polygon", "PBR", "UV mapping"
- **Material Detail**: Specific about textures needed
- **Export Formats**: FBX for Unity, OBJ for Blender
- **Use Case**: Populate 3D property scenes, product visualization
- **Integration Path**: 
  - Generate 3D model → Import to Blender
  - Apply materials → Place in property scene
  - Render final marketing images
- **Professional Standards**: "architectural visualization quality"

---

## 🔍 Part 3: Critique & Iteration Workflow (30 min)

### Sample Image Critique

**Generated Image Analysis:**

**Image 1: Luxury Villa Exterior**
```
✅ STRENGTHS:
- Composition follows rule of thirds
- Lighting captures golden hour effectively
- Pool water has decent reflections

❌ AREAS FOR IMPROVEMENT:
1. Pool reflections lack sharpness
   → FIX: Add "crystal-clear water reflections, mirror-like surface"
   
2. Furniture appears generic
   → FIX: Specify "designer outdoor furniture, teak loungers"
   
3. Sky could be more dramatic
   → FIX: Add "dramatic clouds, vibrant sunset colors"
   
4. Foreground lacks detail
   → FIX: Include "landscaped garden in foreground, tropical flowers"
```

**Refined Prompt (Version 2):**
```
Ultra-modern luxury villa at golden hour, infinity pool with crystal-clear 
mirror-like reflections, designer teak outdoor loungers, dramatic clouds 
with vibrant sunset colors, landscaped garden in foreground with tropical 
flowers, sleek glass facade, photorealistic, 8K, architectural photography
```

---

### Rapid Iteration Workflow

```
ITERATION PROCESS (Per Asset Type):

┌─────────────────────────────────────────┐
│ Step 1: BATCH GENERATION                │
│ - Generate 3-4 variants of prompt       │
│ - Use same settings for consistency      │
│ - Takes ~2-3 minutes                     │
└───────────────┬─────────────────────────┘
                │
┌───────────────▼─────────────────────────┐
│ Step 2: QUICK SHORTLIST                 │
│ - Team picks top 2 candidates           │
│ - Identify what works vs. doesn't       │
│ - Takes ~2 minutes                       │
└───────────────┬─────────────────────────┘
                │
┌───────────────▼─────────────────────────┐
│ Step 3: GATHER FEEDBACK                 │
│ - What's missing?                       │
│ - What's wrong compositionally?         │
│ - What matches brand guidelines?        │
│ - Takes ~3 minutes                       │
└───────────────┬─────────────────────────┘
                │
┌───────────────▼─────────────────────────┐
│ Step 4: REFINE PROMPT                   │
│ - Add specific missing elements         │
│ - Adjust technical parameters           │
│ - Strengthen quality keywords           │
│ - Takes ~2 minutes                       │
└───────────────┬─────────────────────────┘
                │
┌───────────────▼─────────────────────────┐
│ Step 5: RE-GENERATE                     │
│ - Generate 2-3 refined versions         │
│ - Compare with original batch           │
│ - Takes ~2 minutes                       │
└───────────────┬─────────────────────────┘
                │
┌───────────────▼─────────────────────────┐
│ Step 6: FINAL SELECTION                 │
│ - Choose winner or repeat cycle         │
│ - Maximum 3 iteration cycles            │
│ - Total time: ~15 minutes per asset     │
└─────────────────────────────────────────┘

SCALING STRATEGY:
- Parallel processing: Multiple team members work on different assets
- Template reuse: Successful prompts become templates
- A/B testing: Generate variants of successful prompts for future use
- Feedback database: Track what works for different property types
```

---

## 🔧 Part 4: Integration & Pipeline Proposal (30 min)

### Complete Production Pipeline

```python
# PIPELINE PSEUDOCODE

class CreativeCampaignPipeline:
    """
    End-to-end workflow from concept to campaign-ready visuals
    """
    
    def __init__(self, project_name, creative_brief):
        self.project = self.create_project(project_name, creative_brief)
        self.assets = []
    
    def execute_full_pipeline(self):
        """
        Main pipeline execution
        """
        # PHASE 1: IDEATION
        prompts = self.ideation_phase()
        
        # PHASE 2: GENERATION
        raw_assets = self.generation_phase(prompts)
        
        # PHASE 3: REVIEW & ITERATION
        refined_assets = self.review_and_iterate(raw_assets)
        
        # PHASE 4: INTEGRATION & EXPORT
        final_assets = self.integration_phase(refined_assets)
        
        # PHASE 5: DELIVERY
        self.campaign_launch(final_assets)
        
        return final_assets
    
    # ============================================
    # PHASE 1: IDEATION
    # ============================================
    def ideation_phase(self):
        """
        Creative team brainstorms and refines prompts
        """
        prompts = {
            'images': [],
            'videos': [],
            '3d_assets': []
        }
        
        # Load templates based on project type
        if self.project.type == "luxury_real_estate":
            prompts['images'] = [
                self.load_template("villa_exterior"),
                self.load_template("penthouse_interior"),
                self.load_template("amenity_spaces")
            ]
            prompts['videos'] = [
                self.load_template("aerial_tour"),
                self.load_template("lifestyle_montage")
            ]
            prompts['3d_assets'] = [
                self.load_template("outdoor_furniture"),
                self.load_template("interior_decor")
            ]
        
        # Team adds custom refinements
        prompts = self.team_refinement_session(prompts)
        
        # Version control
        self.save_prompt_versions(prompts)
        
        return prompts
    
    # ============================================
    # PHASE 2: GENERATION
    # ============================================
    def generation_phase(self, prompts):
        """
        Generate all asset types using appropriate AI tools
        """
        assets = []
        
        # Generate Images
        for img_prompt in prompts['images']:
            result = self.generate_image(
                prompt=img_prompt['text'],
                provider="huggingface",
                model="flux-schnell",
                n_variants=3  # Generate multiple for selection
            )
            assets.append({
                'type': 'image',
                'prompt': img_prompt,
                'outputs': result,
                'status': 'pending_review'
            })
        
        # Generate Videos
        for vid_prompt in prompts['videos']:
            result = self.generate_video(
                prompt=vid_prompt['text'],
                provider="runwayml",
                duration=10,
                fps=30
            )
            assets.append({
                'type': 'video',
                'prompt': vid_prompt,
                'outputs': result,
                'status': 'pending_review'
            })
        
        # Generate 3D Assets
        for asset_prompt in prompts['3d_assets']:
            result = self.generate_3d_model(
                prompt=asset_prompt['text'],
                provider="shap_e",
                export_format="fbx"
            )
            assets.append({
                'type': '3d',
                'prompt': asset_prompt,
                'outputs': result,
                'status': 'pending_review'
            })
        
        return assets
    
    # ============================================
    # PHASE 3: REVIEW & ITERATION
    # ============================================
    def review_and_iterate(self, assets, max_iterations=3):
        """
        Team reviews assets and iterates until approved
        """
        refined_assets = []
        
        for asset in assets:
            iteration_count = 0
            approved = False
            
            while not approved and iteration_count < max_iterations:
                # Present to team
                feedback = self.get_team_feedback(asset)
                
                if feedback['approved']:
                    asset['status'] = 'approved'
                    refined_assets.append(asset)
                    approved = True
                else:
                    # Refine prompt based on feedback
                    refined_prompt = self.refine_prompt(
                        asset['prompt'],
                        feedback['suggestions']
                    )
                    
                    # Re-generate
                    if asset['type'] == 'image':
                        new_output = self.generate_image(refined_prompt)
                    elif asset['type'] == 'video':
                        new_output = self.generate_video(refined_prompt)
                    elif asset['type'] == '3d':
                        new_output = self.generate_3d_model(refined_prompt)
                    
                    asset['outputs'] = new_output
                    asset['iteration'] = iteration_count + 1
                    iteration_count += 1
            
            if not approved:
                # Mark for manual review if iterations exhausted
                asset['status'] = 'needs_manual_review'
                refined_assets.append(asset)
        
        return refined_assets
    
    # ============================================
    # PHASE 4: INTEGRATION & EXPORT
    # ============================================
    def integration_phase(self, assets):
        """
        Export and integrate assets into production tools
        """
        final_assets = []
        
        for asset in assets:
            if asset['status'] != 'approved':
                continue
            
            if asset['type'] == 'image':
                # Optimize for web
                optimized = self.optimize_for_web(asset['outputs'])
                final_assets.append({
                    'asset': optimized,
                    'destination': 'marketing_materials',
                    'formats': ['jpg', 'png', 'webp']
                })
            
            elif asset['type'] == 'video':
                # Encode for distribution
                encoded = self.encode_video(asset['outputs'])
                final_assets.append({
                    'asset': encoded,
                    'destination': 'campaign_distribution',
                    'formats': ['mp4', 'webm']
                })
            
            elif asset['type'] == '3d':
                # Export to Blender and Unity
                blender_export = self.export_to_blender(asset['outputs'])
                unity_export = self.export_to_unity(asset['outputs'])
                
                final_assets.append({
                    'asset': asset['outputs'],
                    'blender_file': blender_export,
                    'unity_asset': unity_export,
                    'destination': '3d_pipeline',
                    'formats': ['fbx', 'obj', 'blend']
                })
        
        return final_assets
    
    # ============================================
    # PHASE 5: CAMPAIGN LAUNCH
    # ============================================
    def campaign_launch(self, final_assets):
        """
        Final review and campaign deployment
        """
        # Create campaign package
        package = {
            'project_id': self.project.id,
            'assets': final_assets,
            'metadata': self.generate_metadata(),
            'launch_date': datetime.now()
        }
        
        # Save to library
        self.save_to_project_library(package)
        
        # Generate delivery report
        report = self.generate_delivery_report(package)
        
        return package, report
    
    # ============================================
    # HELPER METHODS
    # ============================================
    def generate_image(self, prompt, provider="huggingface", **kwargs):
        """Call image generation API"""
        api = self.get_api_client(provider)
        return api.generate(prompt, **kwargs)
    
    def export_to_blender(self, model_path):
        """
        Export 3D asset to Blender format
        """
        # Python script that runs in Blender
        blender_script = f"""
import bpy

# Import the FBX file
bpy.ops.import_scene.fbx(filepath="{model_path}")

# Set up materials and lighting
# ... material setup code ...

# Save as .blend file
output_path = "{model_path.replace('.fbx', '.blend')}"
bpy.ops.wm.save_as_mainfile(filepath=output_path)
"""
        # Execute script (would need Blender CLI or API)
        return self.execute_blender_script(blender_script)
    
    def export_to_unity(self, model_path):
        """
        Prepare asset for Unity import
        """
        # Ensure FBX is Unity-compatible
        # Set proper scale, rotation, materials
        unity_ready_path = self.convert_for_unity(model_path)
        return unity_ready_path


# ============================================
# EXAMPLE USAGE
# ============================================
if __name__ == "__main__":
    # Initialize pipeline
    brief = {
        'project_name': 'Luxury Beachfront Villas Q4 2024',
        'description': 'Marketing campaign for ultra-luxury beachfront properties',
        'target_audience': 'Ultra-high-net-worth individuals, 45-65 years old',
        'style_guidelines': 'Photorealistic, warm tones, lifestyle-focused',
        'deliverables': ['Hero images', 'Video tour', '3D walkthrough assets']
    }
    
    pipeline = CreativeCampaignPipeline(
        project_name=brief['project_name'],
        creative_brief=brief
    )
    
    # Execute full workflow
    final_assets = pipeline.execute_full_pipeline()
    
    print(f"Campaign complete! Generated {len(final_assets)} final assets.")
```

---

## 🚀 Presentation Structure (30 min)

### Slide Flow

1. **Opening (2 min)**
   - Problem statement
   - Solution overview

2. **System Architecture (8 min)**
   - Show architecture diagram
   - Explain module interactions
   - Highlight integration points

3. **Live Demo (10 min)**
   - Walk through Streamlit interface
   - Generate sample images
   - Show feedback system
   - Demonstrate export features

4. **Prompt Engineering Examples (5 min)**
   - Show 5 luxury real estate prompts
   - Explain strategy for each

5. **Iteration Workflow (3 min)**
   - Critique sample outputs
   - Show refinement process

6. **Pipeline & Scaling (2 min)**
   - Present pseudocode
   - Discuss production readiness

7. **Q&A**

---

## ✅ Checklist for Completion

- [ ] System architecture diagram created
- [ ] Streamlit app with all 7 tabs functional
- [ ] 5 prompts written with strategies
- [ ] Image critique examples prepared
- [ ] Pipeline pseudocode documented
- [ ] Blender/Unity integration outlined
- [ ] Demo data populated
- [ ] Presentation slides ready
- [ ] Speaking points prepared

---

## 💡 Tips for Presentation

1. **Start with live demo** - Show don't tell
2. **Emphasize workflow** - Not just tech
3. **Show iteration process** - Critical for buy-in
4. **Highlight scalability** - Growth potential
5. **Be honest about limitations** - Demo vs production
6. **Connect to business value** - ROI, efficiency gains

---

## 📊 Expected Outcomes

By completing this assignment you demonstrate:

✅ Understanding of AI tool landscape  
✅ Prompt engineering expertise  
✅ System design capabilities  
✅ Integration knowledge (Blender/Unity)  
✅ Workflow optimization skills  
✅ Production readiness mindset

Good luck! 🚀