MCP Veo 3 Video Generation Server

# Veo 3 MCP Server Examples This directory contains example scripts demonstrating how to use the Veo 3 MCP server. ## Prerequisites 1. **Set up your API key**: ```bash export GEMINI_API_KEY="your_gemini_api_key_here" ``` 2. **Install dependencies**: ```bash cd .. pip install -r requirements.txt ``` ## Example Scripts ### 1. `test_basic_generation.py` Tests basic text-to-video generation capabilities. **Features tested:** - Simple text-to-video generation - Video generation with negative prompts - Different model options **Run:** ```bash python test_basic_generation.py ``` ### 2. `test_image_to_video.py` Tests image-to-video generation capabilities. **Features tested:** - Creating sample images for testing - Image-to-video generation - Working with existing image files **Run:** ```bash python test_image_to_video.py ``` ### 3. `test_mcp_client.py` Tests the MCP server through proper MCP protocol communication. **Features tested:** - MCP server initialization - Tool listing - Video generation through MCP calls - Video file management **Run:** ```bash python test_mcp_client.py ``` ### 4. `test_fastmcp_server.py` Tests the updated FastMCP-based server functionality. **Features tested:** - Server startup and dependency checks - Configuration validation - FastMCP framework compatibility **Run:** ```bash python test_fastmcp_server.py ``` ### 5. `test_uv_veo3.py` (New) Tests uv and uvx integration for the Veo 3 server. **Features tested:** - uv run integration (development) - uvx integration (after PyPI publication) - Cursor-style MCP client usage **Run:** ```bash python test_uv_veo3.py ``` ## Expected Output ### Successful Generation ``` ✅ Video generated successfully! 📁 File: test_videos/veo3_video_20240115_143022.mp4 🎬 Model: veo-3.0-fast-generate-preview ⏱️ Generation time: 45.3 seconds 📏 File size: 12.4 MB ``` ### Common Issues 1. **API Key Missing**: ``` ❌ Error: GEMINI_API_KEY environment variable not set ``` **Solution**: Set your Gemini API key as shown above. 2. **Generation Timeout**: ``` ❌ Video generation timed out after 600 seconds ``` **Solution**: Try again during off-peak hours or use a simpler prompt. 3. **Invalid Image Path**: ``` ❌ Error: Image file not found: /path/to/image.jpg ``` **Solution**: Check that the image file exists and the path is correct. ## Tips for Testing ### Effective Test Prompts **Short and Simple** (faster generation): - "A gentle ocean wave on a beach" - "Clouds moving across a blue sky" - "A flower swaying in the breeze" **Detailed Cinematic** (higher quality): - "A tracking shot of a red car driving through a desert landscape at golden hour" - "Close-up of raindrops falling on a window with soft lighting" ### Model Selection for Testing - **veo-3.0-fast-generate-preview**: Use for quick testing (faster) - **veo-3.0-generate-preview**: Use for final/production videos (higher quality) - **veo-2.0-generate-001**: Use for comparison or when audio isn't needed ### Monitoring Generation The scripts provide progress updates: - Initial request confirmation - Polling status every 10 seconds - Final results with timing and file info ## Troubleshooting ### Server Won't Start ```bash # Check if all dependencies are installed pip list | grep -E "(mcp|google-genai)" # Try running the server directly python ../mcp_veo3.py ``` ### Generation Takes Too Long - Try simpler prompts - Use `veo-3.0-fast-generate-preview` model - Test during off-peak hours - Check your internet connection ### File Permissions ```bash # Ensure the examples directory is writable chmod +w . mkdir -p test_videos ``` ## Output Files Generated videos are saved in: - `test_videos/` - For basic generation tests - `mcp_test_videos/` - For MCP client tests Files are named with timestamps: `veo3_video_YYYYMMDD_HHMMSS.mp4` ## Next Steps After running these examples successfully: 1. **Integrate with your MCP client** using the configuration in the main README 2. **Experiment with different prompts** to understand Veo 3's capabilities 3. **Try image-to-video** with your own images 4. **Explore different models** for various use cases