💡 Inspiration

We all love watching series and movies, but life gets busy. We miss episodes, forget the plot after a long break, or don't have 2 hours to watch a full movie just to know what happens.

Fire TV is built for the big screen, and Vega OS is Amazon's fast and lightweight platform for it. We asked: what if your TV could tell you the story so far, made just for you?

So we built STORYICUE, a Fire TV app that gives a personalized recap and summary with the help of AI for any video you play. You get the core of the story quickly, without watching everything.

🎯 What It Does

Shows a personalized recap ("Previously on...") of the series or video you are watching
Gives a short summary with key characters, plot points and important moments
Lets you choose the summary length (quick or detailed)
Works fully with the TV remote
Built for the big screen, with large text and simple navigation

🛠️ How We Built It

STORYICUE is a native Vega OS app for Fire TV, built with React Native and TypeScript.

Layer	Technology
Platform	Vega OS, Vega SDK, Kepler CLI, Fire TV (TV target)
Frontend	React Native 0.83, React 19.2, TypeScript 5.8
Vega UI library	@amazon-devices/react-native-kepler
Build tools	Node.js 20+, npm, Metro, Babel, Bash scripts
Testing	Vega Virtual Device, Jest, React Native Testing Library
Code quality	ESLint (with Kepler plugin), Prettier
Input data	[Subtitles / transcripts / metadata]
AI model	[Your AI model or API]
Backend	[Your backend, or "None, runs on device"]

Flow:

Video plays → Get video context → AI creates summary → Personalize recap → Show on TV screen

⏱️ Time Saved

$$ \text{Time Saved} = \left(1 - \frac{T_{\text{recap}}}{T_{\text{video}}}\right) \times 100\% $$

Example: a 45 minute episode summarized in a 2 minute recap saves about $95.6\%$ of viewing time.

📚 What We Learned

How Vega OS works, including its manifest, build and packaging flow
How to build and test apps on the Vega Virtual Device
How to design for the 10-foot TV experience (big text, D-pad navigation, focus control)
How to use AI to summarize video content
How to keep recaps short, clear and spoiler-free
How to write tests and lint rules for a TV app

🚧 Challenges We Faced

Vega OS is new: There are fewer examples and tutorials, so we read the docs carefully and experimented a lot.
Environment setup: Setting up the SDK, Java, Node and the Vega Virtual Device took many tries and custom scripts.
Video playback: We had to install and configure media codecs so video and audio could play on the virtual device.
TV navigation: Remote and focus handling is very different from touch screens.
Getting video context: [Explain how you got subtitles or scene data and what was hard.]
Spoilers: A good recap must not reveal future events, so we controlled the AI output carefully.
Speed: Summaries had to appear quickly on TV hardware.