OpenMontage

OpenMontage is an open-source, agentic video production system licensed under AGPLv3 and hosted on GitHub. It turns any AI coding assistant — Claude Code, Cursor, Copilot, Windsurf, or Codex — into a full video production pipeline that handles research, scripting, asset generation, editing, and final composition. The project is currently in private alpha with a waitlist open for beta access, and the GitHub repository reports over 53,000 stars.

What It Is

OpenMontage is a pipeline-driven video production framework where the AI coding assistant acts as the orchestrator. Rather than generating a single clip from a prompt, it runs a structured, multi-stage production workflow — the same process a real production team follows — automated end to end. The engine is open source, the project format is published as plain files, and all rendering happens locally on the user's machine with no watermarks and no metering of local compute.

The 9-Stage Pipeline

Every production runs through a fixed sequence of stages, each governed by a dedicated "director skill" (a markdown instruction file the agent reads):

• Brief — audience, duration, tone, key messages
• Research — live web search across YouTube, Reddit, news, and academic sources before any script is written
• Script — a full screenplay with voice-over lines and B-roll notes per line
• Scene Plan — a medium chosen per scene (screen capture, motion graphics, generated clip, stock footage)
• The Gate — a human approval checkpoint with a full cost estimate before any paid generation runs
• Narration — voice persona selection with per-sentence note attachment
• Music — mood/BPM-matched track from a curated free catalog or generated
• Compose — brand rules, captions, and local assembly via Remotion or HyperFrames
• Render — 16:9, 1:1, and 9:16 outputs rendered on the user's machine with a provenance report

Zero-Key and Local-First Operation

OpenMontage is designed so users can produce real videos without any paid API keys. Out of the box, the setup provides Piper TTS for offline narration, free stock footage from Archive.org, NASA, Wikimedia Commons, Pexels, Pixabay, and Unsplash, Remotion for React-based animated composition, HyperFrames for HTML/GSAP motion graphics, and FFmpeg for encoding and post-production. Users with a GPU can also unlock free local video generation via WAN 2.1/2.2, Hunyuan, CogVideo, or LTX-Video. Optional paid API keys (fal.ai, ElevenLabs, Runway, Google Veo, Kling, Suno, and others) unlock premium generation capabilities.

Agent-Native Architecture and MCP Toolset

The engine ships as an MCP toolset, meaning it can run as a plugin inside Claude Code, Codex, or OpenCode — appearing as a toolset in the terminal rather than a separate application. The homepage states it also runs in OpenMontage Studio, a desktop production board in private alpha with a beta releasing soon. The architecture is explicitly agent-first: there is no code orchestrator; the AI coding assistant reads YAML pipeline manifests and markdown skill files, calls Python tools, self-reviews outputs, checkpoints state as JSON, and pauses for human approval at every creative gate.

Production Governance and Quality Gates

OpenMontage treats video production like software engineering, with enforced quality gates at every stage:

• Human approval gates are enforced, not suggested — proposal, script, scene plan, generated assets, and publish all pause for sign-off
• Pre-compose validation blocks render if the delivery promise is violated (e.g., a "motion-led" brief with mostly still images)
• Post-render self-review runs ffprobe validation, frame extraction at four positions, audio level analysis, and subtitle checks — the video is not presented if the review fails
• Slideshow risk scoring uses a 6-dimension analysis to prevent "animated PowerPoint" outputs
• Scored provider selection ranks every tool across 7 dimensions (task fit, output quality, control, reliability, cost efficiency, latency, continuity) and logs the decision trail with alternatives considered
• Budget governance includes cost estimation before execution, per-action approval thresholds, and configurable spend caps

Current Status: Private Alpha, Beta Incoming

The project launched on GitHub in March 2026 and, according to the repository metadata, reached over 53,000 stars and was listed as the #1 Repository of the Day on GitHub Trending. The homepage states the project is in private alpha with a hand-picked crew shipping real launch videos, and weekly beta cohorts are described as coming next. OpenMontage Studio (the desktop app) is in private alpha with a beta releasing soon. The repository was last pushed in August 2026 and shows active development across 12 production pipelines, 100+ registered tools, 60+ provider integrations, and 700+ agent skill and production-knowledge files.