calesthio/OpenMontage

The problem

Producing a short video end to end means jumping between a script doc, stock-footage sites, image and video generators, a TTS service, and a nonlinear editor, then re-exporting every time something changes. That loop favors people who already know editing software. OpenMontage instead packages this process as agent-readable pipelines and skill files, so an AI coding assistant orchestrates research, scripting, asset retrieval, editing and final composition from a plain-language request.

What's different

Most "AI video" tools are hosted products that generate a clip from a prompt and leave assembly to you; the open-source alternative is usually a thin wrapper around one generation API. OpenMontage's bet is on breadth and orchestration: a dozen named pipelines, provider adapters to multiple hosted image, video and TTS services as well as free stock archives, FFmpeg and Remotion for the render, and a local Backlot board that turns asset generation into an approval gate with per-asset cost. It also claims real motion-footage editing from open archives rather than only animating stills. These are process claims, not verified output quality; the README is mostly showcase videos and does not document how the pipelines behave on messy inputs.

Who it's for

Solo creators, technical marketers and developer-adjacent teams who are comfortable in a terminal and already pay for Claude Code, Cursor, Copilot or similar assistants. It suits short-form and explainer work where you want a reproducible, inspectable project folder. It does not suit professional editors who need frame-accurate timeline control, teams that cannot send footage or scripts to third-party generation APIs, or anyone expecting a point-and-click web app. If you only need a caption burn-in or a trim, FFmpeg plus a script is cheaper.

Cost of adoption

Python plus Node tooling for the Remotion path, FFmpeg, and at minimum one assistant subscription. Real output depends on paid provider keys; the README's own examples cite costs of roughly $1 to $5 per short but do not explain how quality degrades with free providers. AGPL-3.0 matters if you plan to host this as a service. The project is young, and the README is thin on setup, troubleshooting and limitations, so expect to read the code and the agent guide to get unstuck.

Verdict

OpenMontage is a new project, created in March 2026 and tracked only since October; the +10,478 stars this month and the #13 monthly Python board spot describe a sharp attention spike, not an established track record. It has not appeared on the daily trending board, and its Hacker News presence is negligible (a 7-point thread). The approach is genuinely interesting, but with 209 days of history and a showcase-heavy README, treat it as an experiment: clone it, run a short project with free providers first, and judge the output yourself before committing any workflow or budget.