OpenMontage
title: OpenMontage type: tool created: 2026-06-29 last_updated: 2026-06-29 related: ["radar/tools/FireRedOpenStoryline", "radar/tools/VideoAgent", "radar/techniques/DetectionThenIdentityVideoPipeline"] sources: ["https://github.com/calesthio/OpenMontage"] radar_quadrant: Tools radar_ring: Assess radar_position: inner
OpenMontage is an open-source agentic video production system that turns natural language prompts into finished videos. It operates through any AI coding assistant (Claude Code, Cursor, Copilot, Windsurf, Codex) and handles the full production chain: research, scripting, asset generation, editing, and final composition. The system provides 12 named pipelines, 52 registered tools, and over 500 agent skills. Licensed AGPLv3. Created March 2026; 27,070 stars and 3,002 forks as of June 2026.
The user describes a video in plain language inside their AI coding assistant. OpenMontage's agent decomposes the prompt into a production plan, selects a pipeline, generates or retrieves assets, and renders the final output via Remotion, a code-driven video composition framework built on Node.js and React. Every creative decision surfaces for user approval before asset generation begins. Before delivery, the system runs automated self-review: ffprobe validation, frame sampling, audio level analysis, and subtitle checks.
A distinct capability is the reference-video clone path. Paste a YouTube, TikTok, Short, or Reel URL and the agent analyzes transcript, pacing, scenes, and style, then proposes two to three differentiated concepts alongside honest cost estimates before any generation starts.
Prerequisites are Python 3.10+, FFmpeg, Node.js 18+, and one API key for at least one provider.
OpenMontage covers two broad production modes. The first uses AI-generated assets: images from OpenAI gpt-image-1 or FLUX, video clips from Kling v3 or Google Veo, narration from ElevenLabs or OpenAI TTS, word-level captions via WhisperX, and royalty-free music sourced automatically. The second builds from a real-footage corpus: the agent retrieves actual motion clips from free stock archives and edits them into a timeline rather than animating still images.
An "atelier" (bespoke) mode allows hand-authored scenes: each scene specifies its own asset, transition, and timing. A demo produced a 70-second history video through atelier mode for $0.02 total cost.
Provider selection is scored across seven dimensions with an auditable decision log surfaced to the user.
Three months after creation (March 2026) the repository reached #1 on GitHub Trending with over 27,000 stars. Working demos span animated shorts (Pixar-style, Ghibli-style), cinematic sci-fi trailers, product ads, and documentary montages, each produced at costs between $0.02 and $1.33. As of June 2026 there are 122 open issues, consistent with a project under rapid public development.
The AGPLv3 license requires that any network-served derivative work release its source code, which constrains proprietary integration.
Placed in Tools / Assess / inner.
The problem OpenMontage solves is video production from scratch: a user describes a video and receives a rendered file, without managing individual tool calls or editing timelines manually. This is a meaningfully different problem from the other video tools on the radar. FireRed-OpenStoryline operates on existing footage through an editing agent loop; VideoAgent focuses on video understanding and semantic remaking. OpenMontage is upstream of both: it produces the raw video that editing tools would later receive.
Inner position within Assess reflects genuine traction (27K stars, working multi-provider demos, active maintenance) and direct domain fit with the video pipeline. The ring stays at Assess because no first-person production use is confirmed and the 122 open issues indicate active early-stage churn.
Trial gate: a finished video produced end-to-end through OpenMontage's agent pipeline and incorporated into the user's own production workflow, with at least one provider API call handled by the agent rather than manually.
