calesthio/OpenMontage: A Claude Code Plugin for Video Production

What is calesthio/OpenMontage

calesthio/OpenMontage (52,955 GitHub stars) is an open source video production plugin. Its own project summary claims 12 production pipelines, 100+ tools, and 700+ supporting skill and production knowledge files, a description of everything bundled inside the repository, not just what surfaces as an invokable Claude Code component.

Flowy indexes a narrower, more concrete slice: 57 components a session can actually invoke, split into 5 agents, 49 skills, and 3 commands. This guide groups those 57 by job, using each one's own indexed description.

Which skills generate video, voice, and images

Most of the video generation work runs through HeyGen. create-video builds a video from a text prompt without picking exact avatars or voices first. avatar-video is the deliberate version, using HeyGen's v2 API to choose an avatar, voice, script, scene, and background for precise control. faceswap swaps a face into an existing video, and video-translate translates and dubs an existing video into another language, with or without lip sync. heygen itself is marked deprecated in favor of those two newer, more focused skills. ltx2 generates video clips from text or an image with the LTX-2.3 22B model, and ai-video-gen does the same through a choice of provider gateways. bfl-api and flux-best-practices cover FLUX image generation for stills that feed into a video.

Audio is its own family of nine skills. acestep generates background music, vocal tracks, covers, and stem extraction with the ACE-Step 1.5 model, and music does instrumental tracks or songs with lyrics through ElevenLabs' Music API. elevenlabs covers the same ground more generally: voiceovers, sound effects, and music from one integration, for podcasts and games as well as video. sound-effects produces non-speech audio, ambient sounds, cinematic impacts, or UI sounds. speech-to-text and azure-speech-to-text transcribe audio to text, and text-to-speech and azure-text-to-speech generate narration. The Azure versions describe themselves as optional cloud providers, used only when an Azure Speech key is configured, with a local path as the default otherwise. setup-api-key walks through configuring the ElevenLabs key several of these need.

Which skills edit and understand footage you already have

A separate, smaller set works locally instead of calling a generation API. video-download pulls source clips from YouTube and other sites, and video-understand extracts frames and transcribes audio to describe what a video contains. Both are documented as needing no API key. ffmpeg and video-edit handle trimming, resizing, concatenation, and format conversion, and playwright-recording captures a browser session as video, useful as source footage. video_toolkit sits above all of these: an autonomous layer that combines voiceover, image generation, music, talking heads, and Remotion rendering into one finished video.

Which skills build the animation and 3D layer

remotion and remotion-best-practices cover building video programmatically in React, this toolkit's own conventions on top of the wider Remotion framework. framer-motion and lottie-bodymovin apply Disney's 12 animation principles to two different renderers, React components and Lottie files exported from After Effects. manim-composer, manimce-best-practices, and manimgl-best-practices cover Manim, the animation library behind 3Blue1Brown-style math and explainer videos: composing a video from a rough idea, and the differences between the Community Edition and the original ManimGL.

Ten more skills cover Three.js on their own. threejs-fundamentals and threejs-animation handle scene setup and keyframe animation, plus more skills for geometry, interaction, lighting, loaders, materials, postprocessing, shaders, and textures, for anyone building or styling a 3D scene in the browser. visual-style sits alongside all of it: it extracts or builds a portable visual design system and applies it to HeyGen videos so a production keeps one consistent look.

What do the three commands do

backlot opens a browser based storyboard board that shows pipeline stages, script, scene plan, and generated assets live as a production runs. animated-drawing takes a supplied photo or drawing and animates it with motion capture into a raster GIF or MP4. ink-art does the opposite: it draws a vector doodle from scratch and animates it, deterministically, into an MP4, either a character walking, dancing, or waving, or a deadpan explanation of a contraption.

Is OpenMontage only for video

No. Five agent components, agent-configuration, client-tools, installation, outbound-calls, and widget-embedding, plus a skill called agents, cover building and deploying ElevenLabs conversational voice agents: phone calls, website widgets, and custom tool integrations. That is a separate use case from producing video. Six more skills, tailwind-design-system, vercel-composition-patterns, vercel-react-best-practices, web-design-guidelines, beautiful-mermaid, and d3-viz, are general web development and diagramming references with nothing video specific in their own descriptions.

How do I install it

Flowy indexes calesthio/OpenMontage from its public GitHub repository. Install it from the listing page linked at the top of this guide, and all 57 components above become available in that session.