Odyssey

Odyssey is a browser interface for interacting with real-time AI video generation. You type a prompt, and a live video stream starts generating in your browser via WebRTC. No page reloads, no waiting for renders — it's a continuous stream that you can steer with new prompts.

The entire app is a single HTML file. No build process, no framework, no node_modules. The Odyssey SDK is imported at runtime from esm.sh. Deploy anywhere that serves static files. This was a deliberate choice — when you're building a tool for rapid experimentation with AI video models, the last thing you want is a 30-second build step between iterations.

The split-panel layout shows your prompt history on the left and the live video stream on the right. You can drag and drop prompts to build visual scripts — sequences of prompts that execute in order, like a storyboard that plays itself. There's a command palette (Cmd+K), prompt history, auto-reconnection when the WebRTC stream drops, and a recordings browser.

The hardest part wasn't the AI or the streaming — it was making the interface feel responsive when the underlying generation has inherent latency. Optimistic UI updates, smooth transitions between prompts, and visual feedback during generation make the difference between a tool that feels like a demo and one that feels like a product.