2014-01-01 project ios life2film video ai montage flagship rust wasm

Life2Film

AI platform that automatically creates beautiful films from your everyday videos — selecting worthy moments, matching music, preserving memories and emotions.

“Start collecting experiences, not money. You are the sum of your experiences… Only your memories and impressions don’t lose their value”

The Problem

Only 1% of people upload to YouTube. Most recorded videos sit unwatched
Storage fills up, memories fade. Children grow up as if they were never small
Creating a nice montage takes hours of editing skill most people don’t have

The Solution

AI watches your videos, finds the best moments, adds music, and creates a film. No editing skills, no equipment, no budget. Minutes, not hours.

Fractal Hypervideo — combine multiple videos into one. Daily films → monthly → yearly. Dive into any moment infinitely deep, like a time machine for your life.

Architecture

Core: video-analyzer — Rust workspace with 11 crates. Frame-level video analysis engine:

Crate	What it does
`va-domain`	Core types, timeline, segments
`va-scene`	Scene detection, shot boundaries
`va-audio`	Audio analysis, beat detection
`va-features`	Visual features, quality scoring
`va-director`	AI montage decisions, story structure
`va-project`	Project management, export
`va-audio-library`	Music matching, mood detection
`va-stt`	Speech-to-text
`va-otio`	OpenTimelineIO integration
`va-wasm`	WebAssembly build for browser
`va-gui-dioxus`	Desktop GUI (Dioxus/Rust)

Platforms (rollout order):

Mobile (first) — iOS beta, then Android
Desktop — Rust native (Dioxus) or Swift (macOS)
Web — WASM build of va-core in browser

Vision

From life2film.com/about:

Life Calendar — 90 years × 52 weeks = 4,680 cells. Each recorded week fills a cell. How many of your weeks have you preserved?
Collaborative editing — group vacation? Everyone uploads, AI creates one combined film
LifeGO — geotag gamification: filming in rare locations yields higher rewards
Creator ecosystem — Creator, Director, Editor, Musician, Sponsor, Moderator roles

Why Rust

The original Life2Film (2014) was Python + Yahoo’s hecate for video analysis. Rewrote it in Rust and got:

100x more convenient — direct FFmpeg API access, frame processing in memory, entire pipeline in one binary
One binary, every platform — compile to iOS, Android, desktop, or WASM for browser. Same core algorithm everywhere
Rust + Swift = Apple must-have — Rust core for performance, Swift for native UI and Apple APIs
DDD finally works — strict types, ownership, algebraic data types. What I was forcing with Pydantic in Python is native in Rust

Cross-platform bridges: WASM for web, PyO3 for Python, UniFFI for Swift/Kotlin/C#.

“Rust — the programming language and my short name (Rust’am). With agents, the more typed and strict the language, the better. Python is for humans — but IDEs aren’t needed anymore.”

The Story

Started in 2014. Built the first version, reached 1M+ users on iOS. Technology wasn’t ready — closed in 2021. The dream never left. SuperDuperAI is the second chapter. Now AI can do what was impossible then: understand video content, detect emotions, match music, create meaningful edits automatically.

Stack

Core: Rust (video-analyzer workspace, 11 crates)
Mobile: Swift (iOS), Kotlin (Android planned)
Web: WASM (va-wasm) + Next.js
Desktop: Dioxus (Rust) or native Swift
AI: on-device where possible (apple-on-device-ai)