What It Is
A whole video team, run by your AI
Most AI video tools hand you one clip from one prompt. OpenMontage hands you an end-to-end production pipeline: research, script, scene plan, assets, edit, and final render. You describe what you want in plain language and your coding agent does the rest, checking in with you at every creative decision.
Here is the part that makes it wild. Most "free AI video" tools quietly mean "animate a few still images." OpenMontage can do that, but it can also build a real video: it pulls actual motion footage from free stock and open archives, ranks it, and cuts it into a real timeline. Free path costs $0. With one API key, cinematic videos run $0.15 to $1.50.
It runs inside any AI coding assistant that can read files and run code: Claude Code, Cursor, Copilot, Windsurf, or Codex.
Step 1
Before you start (4 things)
Python 3.10+
From python.org. The engine that runs the tools.
FFmpeg
brew install ffmpeg on Mac, sudo apt install ffmpeg on Linux. Does the actual video assembly.
Node.js 18+
From nodejs.org. Powers the Remotion render engine.
An AI coding assistant
Claude Code (recommended), Cursor, Copilot, Windsurf, or Codex. This is the brain that runs the show.
Step 2
Install it (2 minutes)
Run these three lines in your terminal:
Terminal
git clone https://github.com/calesthio/OpenMontage.git cd OpenMontage make setup
No make on your machine? Run this one line instead:
Manual setup
pip install -r requirements.txt && cd remotion-composer && npm install && cd .. && pip install piper-tts && cp .env.example .env
On Windows and npm install throws ERR_INVALID_ARG_TYPE? Use npx --yes npm install instead.
Then open the OpenMontage folder in Claude Code and just tell it what you want:
Your first prompt
Make a 60-second animated explainer about how neural networks learn
The Free Path
What you get with zero API keys
make setup alone makes real videos for free:
Narration
Piper TTS, offline and human-sounding.
Real footage
Archive.org, NASA, and Wikimedia Commons (plus Pexels, Unsplash, and Pixabay, whose keys are free to grab).
Animation and render
Remotion (React) and HyperFrames (HTML and GSAP) turn stills into a polished, animated edit.
Captions and post
Auto word-level subtitles, plus FFmpeg for encoding, audio mixing, and color grade.
Two free paths. Image-based video: it generates stills and Remotion animates them. Real-footage documentary: say use real footage only and it cuts actual motion clips from open archives.
Optional
Add keys for more firepower
Every key is optional. Add what you have to your .env file. The high-value ones:
.env
FAL_KEY=your-key # the big unlock: FLUX images + Veo / Kling / MiniMax video OPENAI_API_KEY=your-key # TTS + image generation ELEVENLABS_API_KEY=your-key # premium voices, AI music, sound effects SUNO_API_KEY=your-key # full songs with vocals, any genre PEXELS_API_KEY=your-key # free stock footage (free key)
No surprise bills
Budget is capped (default $10 total) and it pauses for your approval on anything over $0.50. It tells you the cost before it spends.
The Sauce
The best prompts (copy-paste)
Paste any of these into Claude Code after setup. Swap the brackets for your topic.
Start from a video you already love
Usually faster than a blank prompt. It keeps the pacing and hook, changes the topic, and shows you a sample plus a cost estimate first.
Prompt
Here's a YouTube Short I love: [paste link]. Make me something like this, but about [your topic].
Prompt
Analyze this Reel: [paste link]. Give me 3 original variants I could make for my own product launch.
Free, zero keys needed
Works the moment setup finishes. No API keys, no cost.
Prompt
Make a 45-second animated explainer about [topic], with narration and captions.
Prompt
Make a data-driven explainer about [topic] with charts and stat reveals.
Free real footage (the "real video" trick)
The phrase use real footage only is what flips it from animated stills to real motion clips.
Prompt
Make a 90-second documentary montage about [theme]. Use real footage only, no narration, elegiac tone, with music.
With an image or video key (~$0.15 to $1.50)
Add a key like FAL_KEY and it generates the visuals for cents.
Prompt
Create a 30-second Ghibli-style animated video of [scene] at golden hour.
Prompt
Make a product launch teaser for [product name].
Full setup (~$1 to $3)
Multiple keys configured. Cinematic, narrated, scored.
Prompt
Create a cinematic 30-second sci-fi trailer: [concept].
Prompt
Make a 90-second explainer about [topic] for middle-schoolers, with a fun narrator voice and a custom soundtrack.
Pro Tips
The stuff that makes it sing
Reference beats blank prompt
Pasting a TikTok, Reel, Short, or YouTube link gets you a grounded plan faster than describing a video from scratch.
Name the platform and it auto-formats
Shorts, Reels, and TikTok render 9:16. YouTube is 16:9. Instagram Feed is 1:1. Cinematic is 21:9. Just say which one.
It is pipeline-driven (12 of them)
The sleepers: Clip Factory chops one long video into ranked short-form clips, Podcast Repurpose turns audio into video, and Localization and Dub translates and dubs into other languages.
It refuses to ship slideshows
A quality gate blocks "animated PowerPoint" and re-checks the render for black frames, audio levels, and captions before it ever shows you anything.
Have a GPU?
Run make install-gpu to unlock free local video generation (WAN, Hunyuan, CogVideo).