kookaburra-cut

skill
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: Apache-2.0
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 5 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

An AI driven, deterministic animated-video studio for macOS (Apple Silicon).

README.md

Kookaburra Cut app icon

Kookaburra Cut

An AI driven, deterministic animated-video studio for macOS (Apple Silicon).

CI CodeQL OpenSSF Scorecard

Kookaburra Cut renders every visual (animated text, graphics, and 3D) through a single
react-three-fiber WebGL canvas, and exports video by stepping a manual clock
frame-by-frame into a bundled ffmpeg sidecar (H.264 / H.265 / ProRes). It runs in
pure WKWebView via Tauri 2, no Chromium. Two exports of the same project are
byte-identical by design. Scenes can also be authored conversationally: the app
embeds a Claude Code terminal that reads and rewrites your project's scene files.

The Kookaburra Cut studio: editing a device scene while the embedded Claude Code terminal rewrites it

Reworking a device scene from the embedded agent terminal. More at kookaburracut.com.

Download

Download for macOS

  • A signed and notarised .dmg; macOS 13+ on Apple Silicon (arm64) only, no Intel build
  • Free and open source; everything runs locally (no accounts, no telemetry, no cloud)
  • The optional AI authoring path uses your own Claude Code install

Status: early release, in active development. Two opt-in network exceptions
to "fully local": the optional embedded Claude Code terminal, which only ever
runs when you explicitly invoke it (it installs via a script shown in the
terminal and talks to Anthropic while you use it), and the update check, off
until you enable it, which asks GitHub whether a newer release exists and
sends no identifiers.

Core features

  • AI-authored scenes: an embedded Claude Code terminal scoped to the open
    project. The output is plain, editable TSX built on a typed toolkit, not a
    black box.
  • Deterministic export: a manual clock stepped frame-by-frame through one
    WebGL canvas into the ffmpeg sidecar. The same project produces the same
    bytes, every time.
  • Device mockups: video and images play on a studio-lit 3D handset through
    a deterministic frame-extraction pipeline, with per-scene camera orbits.
  • Motion that looks finished: headline presets, animated counters, a
    transition pack with live preview, LUT grades and film grain.
  • One project, every feed: 16:9 / 9:16 / 1:1 / 4:5 from the same scenes,
    plus platform export presets (social, CTV, web, ProRes master) with size
    estimates and loudness targeting.
  • Projects are just files: a folder of project.json + scenes/*.tsx +
    assets. Git-friendly, no opaque project format.
  • Themes: JSON documents covering colour, typography, lighting, staging and
    text-motion defaults; swappable per project or per scene.
  • Fully local: preview and export never leave your Mac.

Building from source

Prerequisites:

  • macOS on Apple Silicon, Xcode command-line tools
  • Node ≥ 20.19, pnpm
  • Rust stable (rustup default stable)
  • ffmpeg on PATH (for the dev sidecar copy)
pnpm install
pnpm setup:ffmpeg     # provision the ffmpeg sidecar (dev copy of system ffmpeg)
pnpm tauri dev        # launch the app (Vite HMR + WKWebView shell)

Other commands:

pnpm build            # tsc --noEmit + vite build
pnpm test             # vitest
pnpm lint             # biome check .

Stack: Tauri 2 · React 19 · react-three-fiber · troika SDF text · anime.js v4 ·
zustand · ffmpeg/ffprobe sidecars. Package manager: pnpm. Full version list
in docs/architecture.md.

Note: the photoreal device model is a licensed third-party asset and is
not included in this repository (see Licensing below). Clones build and
run against a bundled generic placeholder handset; dropping a licensed model
into src/assets/models/licensed/ overrides it (see
src/assets/models/README.md).

Layout

src/toolkit/   shipped authoring primitives  (the @kookaburra/toolkit surface)
src/engine/    timeline · format · compositor · exporter (deterministic loop)
src/theme/     theme schema, bundled themes, fonts
src/store/     zustand editor/preview state
projects/      one folder per project: project.json + scenes/ + assets/
src-tauri/     Rust shell, tauri.conf.json, capabilities/, bin/ (sidecars)
docs/          architecture, determinism, decisions, design, voice
.claude/       project skills (scene authoring, export presets) + slash commands

Authoring scenes

Scenes live in projects/<project>/scenes/*.tsx as defineScene default exports and
use @kookaburra/toolkit only: never the DOM or the wall clock. See the
kookaburra-scene-authoring skill and /new-scene, or
docs/determinism.md for the why.

Documentation

Start at docs/README.md. The short version:
docs/architecture.md explains how rendering and export
work, docs/determinism.md is the byte-identical-export
contract, and docs/decisions.md records every locked
decision and why.

Contributing & security

This is a personal, source-available project. Issues and discussion are welcome;
pull requests may be declined: see CONTRIBUTING.md.
Security reports: see SECURITY.md.

Licensing

Kookaburra Cut's code is dual-licensed under either the
MIT License or the Apache License 2.0, at
your option. Bundled assets keep their own terms: fonts are OFL 1.1, HDRI
environments are CC0 (Poly Haven), and the photoreal device model is a
purchased licensed asset that is deliberately not in the repo (its
committed preview images are not reusable outside this project). The ffmpeg
sidecar the app drives is a GPL build (libx264/libx265), provisioned
locally by script and never committed here. Full inventory:
NOTICE.md.

Trademarks & affiliation

Kookaburra Cut is an independent project. Apple and iPhone are trademarks of
Apple Inc.; device names in the catalogue identify the hardware being
depicted. Platform names in export presets (Meta, TikTok, YouTube, LinkedIn,
X, Reddit, Telegram) identify export targets only. This project is not
affiliated with, sponsored by, or endorsed by Apple, OpenAI, Anthropic, or
any of those platforms.

Yorumlar (0)

Sonuc bulunamadi