open-montage

name	open-montage
description	OpenMontage™ — AI-powered end-to-end video production. Trigger when user says things like: "make a video", "create a montage", "produce a trailer", "edit this footage", "animated explainer", "podcast clip", "documentary", "cinematic trailer", "talking head video", "repurpose this content", "create a video from", "video about".
version	1.0.0
author	Pauli Second Brain™ | Kupuri Media™
invocation	{"model":"default"}
metadata	{"hermes":{"tags":["video","production","filmmaking","montage","editing","media","AI","animation"],"related_skills":[]}}

OpenMontage™

Describe your vision. AI handles the rest — research, scripting, assets, editing, composition.

When to Activate

Activate this skill when the user:

• Says "make a video" or "create a montage"
• Asks for a cinematic trailer, animated explainer, or documentary
• Wants to repurpose a podcast, interview, or existing footage
• Mentions talking head, B-roll, or video editing
• Asks to analyze a reference video and create a variant
• Says "video about" followed by any topic

12 Production Pipelines

Pipeline	Use When
animated_explainer	Concept or product needs visual explanation
cinematic_trailer	Hype/promo for product, event, or project
documentary_montage	Archive footage + narration storytelling
talking_head	Single-speaker scripted or interview format
podcast_repurpose	Convert audio podcast to video clip
social_short	15–60s vertical or square social content
brand_story	Company/founder narrative
tutorial_walkthrough	Step-by-step instructional
news_recap	Research-driven news summary video
music_video	AI-generated visual + audio sync
product_showcase	E-commerce or SaaS product demo
event_highlight	Conference, meetup, or launch recap

Core Workflow

New video from description

1. Ask user: topic, duration target, style, platform (YouTube / Instagram / LinkedIn)
2. Select pipeline based on format
3. Run python -m open_montage.run --pipeline <name> --prompt "<description>"
4. Report: budget estimate → approval → render → deliver

Reference-driven variant

1. User provides a video URL or file path
2. Run analysis: python -m open_montage.analyze --input <path>
3. Propose 3 differentiated variants
4. User selects → full production run

Repurpose existing content

1. Identify source type (podcast, interview, article)
2. Run python -m open_montage.repurpose --source <path> --format <pipeline>
3. Auto-generate script, B-roll selection, captions

Budget & Approval

• Every run produces a cost estimate before rendering
• Default approval threshold: $5.00 (configurable via MONTAGE_SPEND_CAP)
• Free-tier path available: Archive.org + NASA footage + Piper TTS + local Stable Diffusion

Quality Gates

OpenMontage validates automatically:

• Pre-render: script coherence, asset resolution, audio sync
• Post-render: frame inspection, audio levels, duration check
• Audit trail: full JSON checkpoint of every creative decision

Asset Sources (Free Tier)

• Video: Archive.org, NASA, Wikimedia Commons
• Images: Stable Diffusion (local), Wikimedia Commons
• TTS: Piper TTS (local, zero cost)
• Music: Royalty-free archive

Installation

git clone https://github.com/calesthio/OpenMontage
cd OpenMontage
pip install -r requirements.txt
npm install  # Remotion compositor

Set OPEN_MONTAGE_PATH in ~/.hermes/.env to the install directory.