Somewhere in the last few weeks, a repository showed up on GitHub trending with a claim so specific it almost dared you to test it: your coding assistant can direct, shoot, and edit a video. Not generate a ten-second clip. Produce one — start to finish, for nothing.
If you've spent any time editing video, you know the real cost was never really money — it was the switching. Script in one tab, voiceover tool in another, stock footage site in a third, a timeline editor stitching it all together, and you, the human, doing all the copy-pasting in between. OpenMontage was built by someone who got tired of being that glue. So instead of building another app, they built something stranger: a set of instructions that any AI coding assistant can read and simply carry out.
OpenMontage is a free, open-source framework that turns Claude Code, Cursor, Copilot, Windsurf, or Codex into a working video production studio — research, script, footage, voice, edit, and final render, all from one plain-English brief.
Why "free" is the actual headline
Most AI video tools charge by the second of footage generated, and it adds up fast — a one-minute piece can run you real money before you've even seen a draft. OpenMontage sidesteps that entirely. It's released under an open license, meaning anyone can download it, read every line of it, and run it without a subscription. There's no sign-up wall, no credit system, and no hidden paywall waiting at the export button.
It leans on free public archives — Archive.org, NASA's public footage library, Wikimedia Commons — to pull real motion clips instead of only animating still frames. That single choice is what separates it from a lot of AI video tools: it can produce a documentary-style piece using genuine footage, at zero cost, before you ever touch a paid image or video generation API. If you do want to plug in paid tools like Veo or Kling for more stylised shots, you can — but the framework itself never requires it.
How it actually works, under the hood
There's no dashboard, no proprietary server making decisions behind the scenes. The whole system is built from three plain, readable layers, and your AI assistant is the one operating all of them.
1. Tools
A folder of roughly fifty small Python scripts, each doing one job well — generating an image, synthesising a voice, scoring a music track, analysing a scene. Nothing magical, just well-scoped utilities.
2. Pipelines
Readable YAML files that act like a director's shot list for a specific type of video — a documentary, an explainer, a cinematic trailer. They tell the agent what order things happen in and what "done" looks like at each stage.
3. Skills
Hundreds of markdown files that teach the assistant how to use each tool correctly — the same way a new hire might be handed a production manual on their first day. This is the layer that makes the system easy to inspect, fork, and rewrite, because it's just text.
The agent isn't wrapped around the tool. The agent is the tool.
Getting started, step by step
You won't find a "sign up" button anywhere — this runs locally, on your own machine, through an AI coding assistant you already use or can install for free.
- Install the prerequisites. You'll need
Python,Node.js, andFFmpegon your machine. All three are free downloads, and most setup guides walk through this in under ten minutes. - Pick your AI coding assistant. OpenMontage was built around Claude Code, but it also works with Cursor, Copilot, Windsurf, and Codex. If you don't already have one installed, this is the moment to grab it. Not sure how these autonomous agents stack up, or want a free alternative to something like Devin AI first? This side-by-side breakdown of Devin AI vs. open-source agents is a useful primer before you settle on one.
- Clone the repository. Pull the project down from its GitHub page onto your local machine — the same way you'd clone any open-source codebase.
- Open the project inside your coding assistant. Point Claude Code (or your assistant of choice) at the folder. It will read the pipeline manifests and skill files on its own — no manual configuration required.
- Describe the video you want, in plain English. Something as simple as "make a 60-second explainer about how coral reefs form" is enough. The agent takes it from there — researching, scripting, sourcing footage, narrating, and assembling the timeline.
- Review and refine. Because the whole project lives in readable files, you can ask the agent to redo a scene, swap a music track, or rewrite a line of narration, exactly the way you'd ask a human editor for a revision.
You don't have to start from a blank prompt. OpenMontage can take an existing YouTube video, Short, Reel, or TikTok and reverse-engineer it into a production plan — useful when you want to learn from a style you already admire rather than invent one from scratch.
The twelve production styles it ships with
Each "pipeline" is really a preset — a bundle of pacing, tone, and tooling suited to a particular kind of video, so you're not reinventing structure every time.
How people are actually earning from this
None of this replaces a real production team on a big-budget shoot. But for the kind of video work that fills most of the internet — explainers, social clips, localisation, small-business content — a zero-cost production pipeline changes the math on what's worth taking on.
Freelance video production, without the tool subscriptions
Editors and freelancers normally pay monthly for stock libraries, voice tools, and editing software before they've earned a cent. Running OpenMontage locally removes that overhead, so a freelance video gig on Upwork or Fiverr becomes closer to pure profit, especially for clients who just need a solid explainer or product video.
A faceless YouTube or Shorts channel
The Documentary Montage and Clip Factory pipelines are built for exactly this — narrated videos built from real archival footage, at a volume no single editor could sustain manually. Several creators are already running channels this way, ad revenue and sponsorships doing the earning while the pipeline does the editing.
Local business content, on retainer
Small businesses rarely have a marketing budget for weekly social videos. Offering a "$X a month, four videos delivered" package becomes realistic when your own production cost is close to nothing — your time becomes the only real expense.
Localisation and dubbing work
The dedicated dubbing pipeline re-voices and subtitles existing videos into other languages. For creators or agencies expanding into new markets, this is a service that's usually expensive to outsource — and one you can now offer yourself.
Teaching what you've learned
Because the project is new and moves fast, there's real demand for people who can explain it clearly — paid tutorials, cohort-based courses, or a documented build process shared on Substack or YouTube, monetised the ordinary way content already is.
The honest caveat: OpenMontage automates the mechanical parts of production, not the taste. The videos that actually earn money are still the ones with a clear idea, a decent script, and a creator willing to review and refine what the agent produces rather than publishing the first draft. Treat it as a very fast, very cheap assistant editor — not a replacement for judgment.
Comments
Post a Comment