OpenMontage

// Overview

OpenMontage turns an AI coding assistant into a video production system. Instead of a fixed orchestrator, the agent itself drives the work: the repository supplies executable capabilities (tools and pipeline definitions), operational guidance (skills that teach OpenMontage's conventions) and technical knowledge (provider documentation), and the assistant reads them to plan and execute a production from a plain-English brief such as "make a 60-second animated explainer about how neural networks learn".

Twelve pipelines cover animated explainers, motion graphics, avatar spokespeople, cinematic edits, batch short-form clipping, documentary montage from open archives, hybrid footage, localisation and dubbing, podcast repurposing, screen demos, talking heads and SVG character animation. Each runs the same seven stages — research, proposal, script, scene plan, assets, edit, compose — with human approval gates at the proposal, script, scene-plan, asset and publish steps, plus pre-compose validation and a post-render self-review that inspects the output with ffprobe and frame extraction.

Generation is provider-pluggable: video through Kling, Runway Gen-4, Veo 3.1, MiniMax, Seedance and local WAN 2.2 or Hunyuan; images through FLUX, Imagen, Grok Imagine, GPT Image 2 and Recraft; speech through ElevenLabs, Google TTS, OpenAI TTS and local Piper. Composition happens in Remotion (React) or HyperFrames (HTML/GSAP). A zero-key path exists — Piper for narration, Archive.org, NASA, Wikimedia Commons, Pexels, Unsplash and Pixabay for media, Remotion or HyperFrames plus FFmpeg for assembly — and paid providers run under cost estimation, a configurable spend cap and a per-action approval threshold.

// What it does

• Twelve production pipelines that all follow a research → proposal → script → scene plan → assets → edit → compose flow
• Human approval gates at proposal, script, scene plan, asset and publish, with pre-compose validation and post-render self-review
• Pluggable generation providers for video, image, speech and music, plus local options that need no API key
• Two composition engines: Remotion for React-based rendering and HyperFrames for HTML/GSAP kinetic typography and SVG character rigs
• Budget controls — cost estimates before execution, spend caps, and track-only, warn or hard-limit modes
• Ships assistant-specific configuration for Claude Code, Cursor, GitHub Copilot, Codex and Windsurf over shared AGENT_GUIDE.md and PROJECT_CONTEXT.md

// Getting started

OpenMontage is a repository you clone and then open in your AI coding assistant. It needs Python 3.10+, FFmpeg, Node.js 18+ and one of the supported assistants.

git clone https://github.com/calesthio/OpenMontage.git
cd OpenMontage
make setup

python3 -m venv .venv && source .venv/bin/activate && python -m pip install -r requirements.txt && cd remotion-composer && npm install && cd .. && python -m pip install piper-tts && cp .env.example .env

Make a 60-second animated explainer about how neural networks learn

Commands and code are distilled from the project's own documentation — always check the official repo for the latest.

// When to use it

• Produce an explainer or short-form video end to end from a written brief instead of a timeline
• Batch-extract vertical clips from a long recording with a repeatable pipeline
• Build videos with no paid API keys using Piper, open archives and FFmpeg
• Keep generation spend bounded when a run touches several paid media providers

// How OpenMontage compares

OpenMontage alongside other open-source video generation tools AI/TLDR tracks, ranked by GitHub stars.