Descriptions:
Stephanie Nyarko demonstrates how she built a Claude skill that turns a single product photo into a complete 30-second video advertisement—entirely within Claude Desktop, without opening a separate video editor or the Runway ML interface. The workflow uses Claude’s Cowork feature to access local files and Runway ML’s newly released MCP (Model Context Protocol) connector to dispatch video generation jobs, check their status, and stitch the resulting clips together automatically.
The skill generates a structured storyboard from the product image, breaking the ad into scenes and selecting appropriate Runway ML models per shot—Veo 3.1 for cinematic sequences and Kling for close-up product detail. For a six-scene skincare ad, all six clips are submitted to Runway in parallel, and Claude assembles them into a final output with no manual drag-and-drop required. Nyarko approves each API permission call as it surfaces, but otherwise the agent manages the full pipeline.
Beyond the demo, Nyarko walks through connecting Runway ML to Claude by adding a custom MCP connector via a remote server URL available at Runway’s MCP setup page, making the integration replicable for anyone with a Runway account. The video highlights a broader workflow trend: as MCP connectors become available for creative tools like Runway, complex multi-step media production is increasingly achievable from within a single AI interface rather than across disconnected applications.
📺 Source: Stephanie Nyarko · Published June 06, 2026
🏷️ Format: Hands On Build
