Your AI confidant. Lumis helps you capture the small moments from your day, synthesize research into lessons, and turn your real life into content worth sharing.
Tell Lumis about your life, share research you are interested in, and learn to craft your own stories.
- Define your voice
/initwalks you through setting up your vault, voice, and personalized content toolkit - Captures moments and identifies the "5-second moment," the instant something shifted
- Finds connections to your past moments, things you might not have linked on your own
- Builds a Pattern Map in Obsidian that visualizes how your moments relate over time
- Saves research from articles, books, and videos with TL;DR companions
- Captures inspiration by researching people you admire and connecting them to your vault
- Reads research aloud by converting any vault note to narrated audio with ElevenLabs
- Challenges your thinking through targeted critical thinking prompts with honest feedback
- Develops stories through free writing and guided questions, then shapes them into clean narratives
- Directs video production by turning your story into a shot-by-shot timeline, generating avatar clips, and assembling branded video
- Builds carousels from stories into card-by-card LinkedIn carousel plans with copy and image direction
- Writes articles from stories into full long-form blog posts ready to publish
- Generates images for any director format using Google Imagen, with brand-aware prompts that update source files automatically
- Creates diagrams from stories as interactive React Flow visualizations, with PNG export for embedding anywhere
- Cleans AI writing by detecting and fixing AI vocabulary, filler phrases, em dash overuse, and structural tells
This repo is the engine. Your Obsidian vault is where your moments, stories, and research live. They stay separate so your personal content never ends up in a code repo. Your stories remain private to you.
Lumis owns the full content flywheel. Most tools handle one step. Lumis connects all of them because a video director that doesn't know your moments is useless, and a content tool that doesn't know your voice is just another template.
live your life read something someone inspires you
│ │ │
▼ ▼ ▼
┌─────────┐ ┌─────────┐ ┌───────────┐
│ /moment │ │ /add- │ │ /add- │
│ │ │ research│ │inspiration│
└────┬────┘ └────┬────┘ └─────┬─────┘
│ │ │
│ 5-second moment │ TL;DR companion │ bio researched
│ connections found │ category matched │ vault back-links
│ pattern map built │ │ found
│ │ │
└──────────────────────┼────────────────────────┘
│
┌───────▼───────┐
│ Vault │ moments, stories, research,
│ (Obsidian) │ voice, inspiration, patterns
└───────┬───────┘
│
┌──────────────────┬──┴───────────────┬──────────────────┐
│ │ │ │
▼ ▼ ▼ ▼
┌───────────────┐ ┌───────────────┐ ┌───────────────┐ ┌───────────────┐
│ /craft-content│ │ /craft- │ │ /challenge │ │ /listen │
│ │ │ storytelling │ │ │ │ │
│ free write │ │ │ │ pressure-test │ │ research │
│ find story │ │ practice │ │ your ideas │ │ notes to │
│ shape arc │ │ & develop │ │ honest feedbk │ │ audio (TTS) │
└───────┬───────┘ └───────────────┘ └───────────────┘ └───────┬───────┘
│ │
│ ▼
│ narrated .mp3
┌───▼─────────────────────────────────────────┐
│ Director Cuts │
│ │
│ /director-video shot-by-shot timeline │
│ /director-carousel card-by-card plan │
│ /director-article full blog post │
│ /director-images AI images (Imagen) │
│ /director-diagram React Flow diagram │
└────┬──────────────┬──────────────┬──────────┘
│ │ │
▼ ▼ ▼
timeline.md carousel.md article.md
(shot-by-shot) (LinkedIn) (blog)
│
│ lumis studio render <slug>
▼
┌────▼─────────────────────────────┐
│ Studio │
│ │
│ HeyGen avatar clips │
│ ElevenLabs voiceover audio │
│ Remotion branded render │
└──────────────┬───────────────────┘
│
▼
branded .mp4
(YouTube)
Each step feeds the next. Your moments become the raw material for stories. /craft-content finds the story, and the director skills turn it into platform-ready formats.
git clone https://github.com/allthriveai/lumis.git
cd lumis
npm install
npm run build
claude
/init/init walks you through choosing your vault path, scaffolds the directory structure, interviews you to populate Voice.md, then copies and personalizes the Amplify toolkit (8 hook types, 18 content structures, persuasion glossary) using your voice. Lumis reads Voice.md at every layer to align moments, stories, and video with your identity.
/init Set up vault, voice, and Amplify toolkit
/voice Fill in or redo your Voice.md
/moment Capture a moment
/add-research Save and categorize research
/add-inspiration Capture a person who inspires you
/craft-content Free write, find the story, shape the narrative
/craft-storytelling Practice or develop storytelling from moments
/director-video Build a shot-by-shot timeline and produce video
/director-carousel Build a LinkedIn carousel from a story
/director-article Write a long-form article from a story
/director-images Generate AI images for any director format
/director-diagram Create interactive diagram from a story
/challenge Pressure-test an idea with critical thinking prompts
/listen Convert a research note to narrated audio
/brand Set up your visual brand identity
/humanizer Clean AI-generated writing patterns from text
lumis studio list List all director cuts with status
lumis studio render <slug> Render a director cut to branded video
lumis studio preview Open Remotion Studio for previewing
lumis listen <note> Convert a research note to audio narration
lumis listen --voices List available ElevenLabs voices
lumis capture <cmd> OBS capture (setup, start, stop, list, scene)
- Vault structure — how the Obsidian vault is organized, Voice.md, Amplify toolkit
- Signals — event log connecting pipeline stages
- Memory — session history and user preferences
- MCP Server — all tools and Claude Desktop config
- Studio — video production pipeline, image generation, and API setup
- OBS Capture — OBS integration, screen/camera recording, keyboard shortcuts
- Node.js + TypeScript with ES modules
- Claude API for moment analysis and story development
- gray-matter for YAML frontmatter parsing
- MCP SDK for Claude Code integration
- Remotion for programmatic video rendering
- HeyGen for AI avatar video generation
- ElevenLabs for text-to-speech
- Google Imagen for AI image generation
- React Flow for interactive diagrams (loaded via CDN, no npm dependency)
MIT. See LICENSE.