calesthio/OpenMontage

What it solves

OpenMontage automates the end-to-end video production process, acting as an agentic system that handles research, scripting, asset generation, editing, and final composition. It bridges the gap between a plain-language prompt and a finished video file, whether you want to use AI-generated motion clips, real archival footage, or animated graphics.

How it works

The system is designed to run within AI coding assistants (like Cursor or Claude Code). You provide a prompt, and the agent selects a production pipeline, researches the topic, generates or retrieves assets (images, video, music, voiceover), and uses tools like Remotion or HyperFrames to compose the final video. It features a "Backlot" interface that provides a live, visual storyboard, allowing you to approve assets and script segments before the final render.

Who it’s for

It is for creators, developers, and researchers who want to build video content programmatically or via AI agents. It is particularly useful for those who want to maintain control over the creative process by approving individual scenes and assets while letting the agent handle the technical heavy lifting of assembly and rendering.

Highlights

• Agentic Workflow: Fully automated production pipelines that handle everything from web research to final FFmpeg rendering.
• Flexible Asset Sourcing: Supports both AI-generated video (via providers like Kling, Veo, or local models) and real-world archival footage from sources like Archive.org and NASA.
• Interactive Storyboarding: The "Backlot" dashboard lets you view and approve scene-by-scene progress, including takes and cost estimates, before finalizing.
• Zero-Cost Path: Includes a robust free tier using local TTS (Piper), open archival footage, and web-based rendering engines (Remotion/HyperFrames) for those without paid API keys.
• Extensible Architecture: Built for agentic operation, allowing AI coding assistants to discover and execute specific production skills and tools via a registry.