ai-content-generation

What this skill does

# AI Content Generation

AI-powered content and image generation skill using the content-image-generation MCP server. This skill provides capabilities for generating marketing content with Claude Sonnet 4 or Gemini 2.0 Pro, creating images with Google Imagen 3/4, and generating videos with Google Veo 2/3.

## Core Capabilities

### Image Generation
- Generate images with Google Imagen 3 or Imagen 4
- Support multiple aspect ratios (1:1, 16:9, 4:3, 9:16, custom)
- Quality control (SD or HD)
- Batch generation for multiple images
- Prompt enhancement for better results
- Custom negative prompts for quality control
- Seed-based reproducibility

### Content Generation
- Marketing content with Claude Sonnet 4 or Gemini 2.0 Pro
- Hero section copy and landing page content
- Blog posts and articles
- Product descriptions and feature lists
- Email campaigns and ad copy
- SEO-optimized content
- Tone and style customization

### Video Generation
- Short-form videos with Google Veo 2 or Veo 3
- Multiple aspect ratios and durations
- Prompt-based video creation
- Quality and format options

### Cost Optimization
- Pre-generation cost estimation
- Batch optimization recommendations
- Quality vs cost tradeoffs
- Token usage tracking

## Instructions

### Setup and Validation

Before generating content, verify MCP integration:

```bash
# Validate MCP server configuration
bash scripts/validate-mcp-setup.sh

# Setup environment variables
bash scripts/setup-environment.sh
```

### Image Generation Workflow

1. **Enhance Prompt**: Use prompt enhancement for better results
   ```bash
   bash scripts/enhance-image-prompt.sh "your basic prompt"
   ```

2. **Estimate Cost**: Calculate generation costs
   ```bash
   bash scripts/calculate-cost.sh --type image --quality HD --count 5
   ```

3. **Generate Image**: Use MCP tool with enhanced parameters
   - Read template: `templates/typescript/image-generation.ts` or `templates/python/image-generation.py`
   - Call MCP tool: `generate_image_imagen3` or `batch_generate_images`
   - Save to assets directory

4. **Validate Output**: Check image quality and dimensions
   ```bash
   bash scripts/validate-output.sh --type image --path /path/to/image.png
   ```

### Content Generation Workflow

1. **Define Requirements**: Gather content parameters
   - Content type (hero, blog, product, email)
   - Topic and keywords
   - Tone and style (professional, casual, technical)
   - Target audience
   - Desired length

2. **Estimate Cost**: Calculate generation costs
   ```bash
   bash scripts/calculate-cost.sh --type content --model claude-sonnet-4 --length 1000
   ```

3. **Generate Content**: Use MCP tool
   - Read template: `templates/typescript/content-generation.ts` or `templates/python/content-generation.py`
   - Call MCP tool: `generate_marketing_content`
   - Save to content directory

4. **Review and Refine**: Validate content quality
   ```bash
   bash scripts/validate-output.sh --type content --path /path/to/content.md
   ```

### Video Generation Workflow

1. **Prepare Prompt**: Create detailed video description
   ```bash
   bash scripts/enhance-video-prompt.sh "your video description"
   ```

2. **Estimate Cost**: Calculate video generation costs
   ```bash
   bash scripts/calculate-cost.sh --type video --duration 5 --quality HD
   ```

3. **Generate Video**: Use MCP tool
   - Read template: `templates/typescript/video-generation.ts` or `templates/python/video-generation.py`
   - Call MCP tool: `generate_video_veo3`
   - Save to assets directory

### Batch Operations

For multiple assets, use batch generation:

```bash
# Optimize batch parameters
bash scripts/optimize-batch.sh --type image --count 10 --budget 50

# Generate batch with optimized settings
# Use batch_generate_images MCP tool
```

## MCP Tools Reference

### Image Generation
- `generate_image_imagen3`: Generate single image with Imagen 3/4
  - Parameters: prompt, negative_prompt, aspect_ratio, quality, model, seed
  - Returns: Base64 image data, metadata, cost

- `batch_generate_images`: Generate multiple images efficiently
  - Parameters: prompts array, shared settings, batch_size
  - Returns: Array of images with metadata

### Content Generation
- `generate_marketing_content`: Generate marketing copy
  - Parameters: topic, content_type, tone, style, length, model, keywords
  - Returns: Generated content, metadata, cost

### Video Generation
- `generate_video_veo3`: Generate video with Veo 2/3
  - Parameters: prompt, duration, aspect_ratio, quality, model
  - Returns: Video data, metadata, cost

### Utilities
- `calculate_cost_estimate`: Estimate generation costs
  - Parameters: operation_type, parameters, quantity
  - Returns: Cost breakdown, recommendations

- `image_prompt_enhancer`: Enhance image prompts
  - Parameters: basic_prompt, style, quality_level
  - Returns: Enhanced prompt, suggestions

## Scripts Reference

All scripts are located in `skills/ai-content-generation/scripts/`:

- `setup-environment.sh`: Configure environment variables and credentials
- `validate-mcp-setup.sh`: Verify MCP server connection and tools
- `enhance-image-prompt.sh`: Improve image generation prompts
- `calculate-cost.sh`: Estimate generation costs before execution
- `validate-output.sh`: Check quality of generated assets
- `optimize-batch.sh`: Optimize batch generation parameters
- `test-generation.sh`: Run test generation to verify setup

## Templates Reference

### TypeScript Templates (`templates/typescript/`)
- `image-generation.ts`: Complete image generation implementation
- `content-generation.ts`: Marketing content generation
- `video-generation.ts`: Video generation workflow

### Python Templates (`templates/python/`)
- `image-generation.py`: Complete image generation implementation
- `content-generation.py`: Marketing content generation
- `video-generation.py`: Video generation workflow

## Examples

See `examples/` directory for comprehensive usage examples:

- `basic-usage.md`: Simple image and content generation
- `advanced-usage.md`: Batch operations, cost optimization, custom parameters
- `common-patterns.md`: Hero images, blog headers, product galleries
- `error-handling.md`: Retry logic, fallbacks, validation
- `integration.md`: Astro integration, asset management, workflow automation

## Best Practices

### Image Generation
- Always enhance prompts for better quality
- Use HD quality for hero sections and key visuals
- Use SD quality for thumbnails and secondary images
- Specify negative prompts to avoid unwanted elements
- Use consistent seeds for reproducible results
- Batch similar images to optimize costs

### Content Generation
- Provide clear topic and keywords
- Specify target audience for better relevance
- Choose appropriate tone and style
- Review and customize generated content
- Use Claude for technical/detailed content
- Use Gemini for creative/marketing content

### Cost Optimization
- Estimate costs before generation
- Use batch operations for multiple assets
- Choose SD quality when HD is not required
- Optimize prompt length for content generation
- Cache and reuse similar assets
- Monitor token usage and costs

### Error Handling
- Validate MCP setup before operations
- Check API quotas and limits
- Implement retry logic for transient failures
- Validate output quality after generation
- Log costs and metadata for tracking

## Requirements

### Environment Variables
- `GOOGLE_CLOUD_PROJECT`: Google Cloud project ID for Vertex AI
- `ANTHROPIC_API_KEY`: API key for Claude Sonnet content generation
- `GOOGLE_AI_API_KEY`: (Optional) API key for Gemini content generation

### MCP Configuration
- content-image-generation MCP server must be configured in `.mcp.json`
- Google Cloud credentials must be set up for Vertex AI
- Appropriate APIs must be enabled (Vertex AI, Imagen, Veo)

### Project Structure
- Assets directory for storing generated images/videos
- Content directory for storing generated text
- Environment file for API credentials