OpenMontage is an open-source project on GitHub (calesthio/OpenMontage) that calls itself an "agentic video production system". There is no separate app: you open the repository in an AI coding assistant — Claude Code, Cursor, GitHub Copilot, Windsurf or Codex, each with its own instruction file — describe the video you want, and the assistant acts as the producer. It runs every job through the same stages: research, proposal, script, scene plan, assets, edit and compose.
The README lists pipelines such as Animated Explainer, Documentary Montage, Screen Demo, Talking Head, Podcast Repurpose and Localization & Dub. Videos are assembled with Remotion (React), HyperFrames (HTML) and FFmpeg. Besides animating still images, the agent can pull real footage from free archives such as Archive.org, NASA and Wikimedia Commons and cut it into a timeline.
Cost: every API key is optional. The default setup uses free tools, including Piper for offline narration. With paid image or video providers added, the README estimates about $0.15–$1.50 per video, and one showcase video cost $1.33. A default budget cap of $10 and an approval prompt for any single action over $0.50 guard against surprise bills.
Traction: the repository gained about 4,400 stars this week on GitHub's trending list and shows about 78,600 stars and 10,500 forks. It has been around since mid-2026: explainx.ai reported it reaching #1 on GitHub Trending in June. Hacker News discussion has been small (a few posts of under 10 points).
To try it: git clone https://github.com/calesthio/OpenMontage.git, then cd OpenMontage and make setup, and open the folder in your coding assistant. You need Python 3.10+, FFmpeg and Node.js (the README asks for a patched 22.x, 24.x or 26.x release). Limits: the licence is AGPL-3.0, not MIT; there are no tagged releases yet; local LLM support (Ollama, LM Studio) is marked "coming soon"; free local video generation needs a GPU; and without paid providers the output is built from still images and archive footage.
