OpenMontage#

OpenMontage is an agent-orchestrated video production platform. An LLM coding assistant (Claude Code, Cursor, Copilot, or similar) acts as the control plane: it reads a YAML pipeline manifest to determine stage order, follows Markdown stage-director skills for instructions, calls Python tools via the registry, and writes JSON checkpoints to persist state and artifacts. Python supplies only tools and persistence; there is no separate runtime orchestrator.

The system supports both cloud API providers and local/open-source alternatives for every capability. Tools cover analysis, audio (TTS and music), avatars, enhancement, graphics, subtitles, and video generation/composition. Twelve production pipelines are defined in pipeline_defs/, each declaring stages, required artifacts, available tools, and human-approval defaults.

Who it's for#

OpenMontage is for users who produce videos with the help of an AI coding assistant. It is designed for workflows that range from research-driven explainers and cinematic montages to character animation and screen recordings. The agent makes creative decisions while the tool registry and checkpoints enforce budgets, schema validation, and locked decisions such as render runtime and delivery promise.

Documentation structure#

Start with the Getting Started page for installation, prerequisites (Python 3.10+, FFmpeg, Node.js 18+), environment setup, and your first production run.

The Concepts section explains the underlying systems:

The Guides section covers practical usage:

The Reference section documents the concrete surface:

The Troubleshooting section addresses common issues:

All pages are grounded in the actual manifests, tool contracts, schemas, and runtime behavior present in the repository.