Inspired by the Greek messenger goddess of swift dispatches: personalised, broadcast-ready audio news briefings tailored for your daily commute with Gemini AI.
Iris transforms your morning or evening commute into a curated radio newsroom. Instead of scanning endless headlines or listening to generic podcasts that don't match your travel duration, Iris calibrates your curated reading list to the exact length of your journey, synthesising an articulate, multi-story spoken broadcast with natural British English delivery.
Manage articles, filter by beat (Tech, Climate, Economy, Science), customise commute durations from 5 to 45 minutes, select broadcast formats (Solo Host or Co-Hosted Duo), and choose voices.
Listen to studio-grade broadcasts with scrubbing controls, segment chapter markers, and interactive transcripts that jump playback directly to the corresponding story when clicked.
Iris uses a decoupled, full-stack architecture combining a high-performance React SPA with a Node.js/Express server that acts as a secure proxy to Google Gemini models and Firebase Firestore.
flowchart TB
subgraph Client ["Client Application (React 18 + Vite + Tailwind CSS)"]
UI["User Interface\n• Article Manager\n• Commute Config Panel\n• Listen Later Queue\n• Saved Digests"]
AudioEngine["Web Audio Player Engine\n• HTML5 Audio & Waveforms\n• Synchronised Transcript Scroller\n• Segment Chapter Controller"]
LocalCache["Browser Storage & Local State\n• LocalStorage Caches\n• Audio Blob Object URLs\n• Pending Articles Queue"]
end
subgraph Server ["Full-Stack Proxy Server (Node.js / Express on Port 3000)"]
API["REST API Router\n• /api/generate-summary\n• /api/generate-audio\n• /api/generate-article-audio\n• /api/fetch-url\n• /api/search-topics"]
AudioCompiler["PCM to RIFF/WAV Audio Synthesiser\n• 24kHz Mono 16-bit PCM\n• Dynamic WAV Header Encoding\n• Audio Segment Sequencer"]
end
subgraph Gemini ["Google Gemini AI Engine"]
ScriptModel["gemini-3.8-flash\n• UK English Editorial Structuring\n• Time-Calibrated Script Generation\n• Multi-Speaker Dialogue Formats"]
TTSModel["gemini-3.1-flash-tts-preview\n• Expressive Spoken Audio Synthesis\n• Voice Personas (Puck, Charon, Aoede, Fenrir)\n• Raw PCM Streaming Audio"]
end
subgraph Cloud ["Cloud Persistence (Google Cloud / Firebase)"]
Firestore[("Firebase Firestore\n• Shared Topic Libraries\n• Public Shareable Link IDs\n• Real-Time Synchronisation")]
end
UI -->|1. Submit Articles & Config| API
API -->|2. Request Spoken Script| ScriptModel
ScriptModel -->|3. Return Formatted JSON Script| API
API -->|4. Synthesise Audio Segments| TTSModel
TTSModel -->|5. Raw PCM Audio Chunks| AudioCompiler
AudioCompiler -->|6. Return Base64 WAV Broadcast| AudioEngine
AudioEngine -->|7. Play & Synchronise UI| UI
UI <-->|Sync Shared Libraries & Listen Later| Firestore
LocalCache <--> UI
+-----------------------------------------------------------------------------------+
| CLIENT BROWSER (SPA) |
| |
| [Article Manager] <---> [Commute Controls] <---> [Audio Player & Transcript] |
| | | ^ |
| v v | |
+---------+------------------------+-----------------------------+------------------+
| |
| POST /api/generate-summary | Playback Data:
| Payload: { articles, commuteMinutes, tone, voices } | WAV Audio + Timings
v |
+----------------------------------------------------------------+------------------+
| EXPRESS BACKEND (server.ts) |
| |
| 1. /api/generate-summary |
| --> Dispatches prompt to Gemini 3.8 Flash |
| --> Generates UK English time-calibrated broadcast script |
| |
| 2. /api/generate-audio |
| --> Calls Gemini 3.1 Flash TTS preview per story segment |
| --> Receives raw 24kHz 16-bit mono PCM chunks |
| --> Prepends canonical 44-byte RIFF/WAV header |
| --> Returns base64 Data URL array to client for seamless playback |
+--------------------------+------------------------------+-------------------------+
| |
v v
+------------------------------+ +-------------------------------+
| Google Gemini AI SDK | | Firebase Firestore |
| | | |
| * gemini-3.8-flash | | * Shared Topic Collections |
| * gemini-3.1-flash-tts | | * Commute Configurations |
| (Aoede, Puck, Charon) | | * Saved Digests History |
+------------------------------+ +-------------------------------+
-
Precision Commute Calibration (5–45 minutes)
- Automatically budgets spoken word count (~130–150 words per minute) to fit your train or motorway journey.
- Adjusts depth from concise rapid-fire bullet briefs to deep analytical discussions.
-
Studio Audio & Natural Spoken Delivery
- Powered by
gemini-3.1-flash-tts-preview. - Supports Solo Host or Co-Hosted Duo formats with natural banter and fluid transitions.
- Distinctive voice personas: Puck (engaging lead), Charon (deep analyst), Aoede (warm narrative), Fenrir (authoritative).
- Powered by
-
British English Localisation (en-GB)
- Complete UK English spelling throughout interface copy, notifications, and AI broadcast prompts.
- BBC-style news production cadence and transport terminology (motorway, tube, rail corridor).
-
Listen Later Queue
- Individual story queue for listening to standalone articles on demand.
- In-memory audio caching to prevent redundant syntheses.
-
Cloud Sharing via Firebase Firestore
- Generate instant shareable URLs (
?share=...) for curated reading lists. - One-click import into other commuters' studios.
- Generate instant shareable URLs (
-
Backup & Portability
- One-click JSON backup export for saved article collections.
- Instant paste or file upload import.
| Layer | Technology | Purpose |
|---|---|---|
| Frontend Framework | React 18 + Vite | Component lifecycle and fast compilation |
| Styling & Layout | Tailwind CSS | Editorial aesthetic, custom scrollbars, responsive design |
| Icons & Visuals | Lucide React | Clean, scalable vector iconography |
| Server / Proxy | Express.js + TSX | Type-safe backend proxy preventing API key exposure |
| AI Text Engine | Google Gemini 3.8 Flash (@google/genai) |
News distillation, story sequencing, UK script crafting |
| AI Spoken Engine | Google Gemini 3.1 Flash TTS Preview | Expressive voice synthesis and raw audio generation |
| Database & Sync | Firebase Firestore | Durable cross-device storage for shared libraries |
- Node.js 18+ or Bun
- A valid Google Gemini API Key (
GEMINI_API_KEY)
Create a .env file in the root directory:
GEMINI_API_KEY=your_gemini_api_key_here
PORT=3000# Install dependencies
npm install
# Start development server
npm run dev
# Build for production
npm run build
# Start production server
npm startThe application will be accessible at http://localhost:3000.
| Endpoint | Method | Description |
|---|---|---|
/api/health |
GET |
Health check verifying server and Gemini API configuration |
/api/generate-summary |
POST |
Generates a time-budgeted, multi-segment broadcast script |
/api/generate-audio |
POST |
Converts broadcast script segments into 24kHz RIFF/WAV audio |
/api/generate-article-audio |
POST |
Generates on-demand spoken audio for a single article |
/api/refresh-articles |
POST |
Re-fetches latest breaking headlines and updates pending list |
/api/search-articles |
POST |
Discovers recent news articles for a user-specified query |
/api/extract-url |
POST |
Extracts title, text, and metadata from web articles |
Released under the MIT Licence. Built with Google AI Studio.