📰 VOXBYTE — AI-Powered News Anchor App

🏅 2nd Place (Runner-Up) at TAG25 NUST

VOXBYTE is a fully automated AI news broadcasting web application that delivers real-time news delivered by an AI-powered news anchor. The application combines intelligent web scraping, AI script generation (using Google's Gemini API), text-to-speech synthesis, and AI avatar video generation to create a complete end-to-end news pipeline—requiring minimal human intervention.

🎯 Project Overview

VOXBYTE automates the entire news-to-broadcast workflow:

News Scraping → Automatically fetches latest headlines from Dawn News
Script Generation → Uses Gemini API to write engaging news anchor scripts
Text-to-Speech → Converts scripts to realistic voiceovers using ElevenLabs API
Avatar Video → Generates AI anchor videos using D-ID API
Web Display → Shows news headlines and videos on a modern, responsive frontend
Scheduled Updates → Runs the entire pipeline automatically every hour

🔥 Core Features

✅ Automated News Scraping — Fetches top 5 headlines from Dawn News hourly
✅ AI Script Generation — Gemini API creates natural, engaging news scripts
✅ Realistic Voiceovers — ElevenLabs API synthesizes professional news anchor voice
✅ AI Avatar Videos — D-ID API generates videos of AI anchors delivering news
✅ Scheduled Automation — APScheduler runs the full pipeline every hour
✅ Live Dashboard — Modern web interface with news display and video player
✅ Interactive Headlines — Users can click to expand article summaries
✅ Responsive Design — Works on desktop, tablet, and mobile devices
✅ 3D Background — Spline integration for engaging hero section

🧩 Technology Stack

Component	Technology	Purpose
Backend	Python, Flask	Web server & API endpoints
AI/LLM	Google Gemini API	Script generation from news
Web Scraping	BeautifulSoup, Requests	Fetch headlines from Dawn News
Text-to-Speech	ElevenLabs API	Generate voiceovers
Avatar Video	D-ID API	Create AI anchor videos
Task Scheduling	APScheduler, Schedule	Hourly automation pipeline
Frontend	HTML5, CSS3, Vanilla JavaScript	Web interface
Data Storage	JSON	Store scraped articles
3D Graphics	Spline	Interactive background animation

📁 Project Structure

VoxByte/
├── README.md                          # Original documentation
├── README_COMPREHENSIVE.md            # This file - FULL DOCUMENTATION
│
├── backend/
│   ├── app.py                        # Flask server (PORT 5000)
│   ├── script_generator.py           # Gemini API integration for script writing
│   ├── news_scrapper.py              # BeautifulSoup web scraper (Dawn News)
│   ├── text_to_speech.py             # ElevenLabs API integration
│   ├── generate_video.py             # D-ID API integration for avatars
│   ├── automation.py                 # End-to-end pipeline orchestrator
│   ├── run_pipeline.py               # Alternative pipeline runner
│   ├── requirements.txt              # Python dependencies
│   │
│   ├── news_articles.json            # Latest scraped news (auto-updated)
│   ├── final_script.txt              # Generated news script
│   ├── final_audio.mp3               # Generated voiceover audio
│   │
│   ├── static/
│   │   ├── styles.css               # Frontend styling
│   │   ├── script.js                # Frontend interactivity
│   │   └── video/
│   │       └── anchor.mp4           # Generated AI anchor video
│   │
│   └── templates/
│       └── index.html               # Main web page
│
└── [Generated Files - Created at Runtime]
    ├── generated_script.txt          # Latest script (alternate location)

🏗️ How It Works (System Architecture)

Complete Pipeline Flow

┌──────────────────────────────────────────────────────────────────┐
│                      VOXBYTE FULL PIPELINE                        │
└──────────────────────────────────────────────────────────────────┘

STEP 1: NEWS SCRAPING (Runs Every Hour)
   ├─ news_scrapper.py starts
   ├─ Fetches https://www.dawn.com/latest-news
   ├─ Extracts 5 latest articles with BeautifulSoup
   │   - Headline from <h2> tag
   │   - URL from <a> tag
   │   - Summary from first 5 paragraphs <p>
   ├─ Creates JSON structure with timestamp
   └─ Saves to: backend/news_articles.json

STEP 2: SCRIPT GENERATION (Triggered On-Demand or Hourly)
   ├─ app.py scheduler triggers scheduled_script_generation()
   ├─ Reads backend/news_articles.json
   ├─ Extracts all headlines and summaries
   ├─ Sends to Google Gemini API with custom prompt:
   │   "You're a scriptwriter for a news channel..."
   ├─ Gemini generates engaging 1-2 minute news script
   ├─ Response is natural, spoken-style English
   └─ Saves to: backend/static/generated_script.txt
                backend/final_script.txt

STEP 3: TEXT-TO-SPEECH (Requires Audio Generation)
   ├─ Reads final_script.txt
   ├─ Sends to ElevenLabs API (https://api.elevenlabs.io/v1/text-to-speech)
   ├─ Parameters:
   │   - Voice ID: "VzCzzZS0ff2iL6Izl8fR" (specific trained voice)
   │   - Model: eleven_monolingual_v1
   │   - Stability: 0.5, Similarity Boost: 0.75
   ├─ Receives MP3 audio stream
   └─ Saves to: backend/final_audio.mp3

STEP 4: AI AVATAR VIDEO GENERATION (Requires Video Generation)
   ├─ Reads final_audio.mp3
   ├─ Encodes audio to Base64 format
   ├─ Sends to D-ID API (https://api.d-id.com/talks)
   ├─ Parameters:
   │   - Avatar: "amy" (D-ID public avatar)
   │   - Audio: Base64-encoded MP3
   │   - Config: fluent=true, pad_audio=0.2s
   ├─ Receives talk_id
   ├─ Polls https://api.d-id.com/talks/{talk_id} every 5 seconds
   ├─ Waits for "result_url" (video generation takes ~30-60 seconds)
   ├─ Downloads MP4 video file
   └─ Saves to: backend/static/video/anchor.mp4

STEP 5: WEB DISPLAY (User Access)
   ├─ Browser requests http://localhost:5000
   ├─ Flask serves index.html template
   ├─ JavaScript executes:
   │   - Loads news from hardcoded data (currently)
   │   - Renders articles as expandable list
   │   - Displays timestamp of last update
   │   - Embeds video player for anchor.mp4
   │   - Spline 3D animation loads in background
   └─ User can:
       - Click article headers to expand summaries
       - Watch AI anchor video
       - Click links to full articles on Dawn News

STEP 6: SCHEDULED REPETITION (Every Hour)
   ├─ APScheduler trigger fires
   ├─ Entire pipeline repeats (steps 1-5)
   ├─ Latest news scraped, script generated, video created
   └─ Frontend automatically shows new content on refresh

Key API Endpoints

Endpoint	Method	Purpose	Parameters
`/`	GET	Serve main webpage (index.html)	None
`/latest-news-script`	POST	Generate script from current articles	None (reads from JSON)

Data Flow Diagram

Dawn News Website
       ↓
   [Scraper] → news_articles.json
       ↓
[Script Generator] → Gemini API → final_script.txt
       ↓
[Text-to-Speech] → ElevenLabs API → final_audio.mp3
       ↓
[Video Generator] → D-ID API → anchor.mp4
       ↓
[Web Server] → Browser → User watches video + news

⚙️ Component Breakdown (Detailed)

1. news_scrapper.py — News Data Harvesting

What it does:

Scrapes Dawn News website automatically
Runs on a schedule (hourly)
Extracts the latest 5 news articles

How it works:

1. Connects to https://www.dawn.com/latest-news
2. Parses HTML with BeautifulSoup
3. For each <article> tag:
   - Extracts headline from <h2>
   - Extracts URL from <a href>
   - Fetches article page
   - Extracts first 5 <p> tags as summary
4. Stores in JSON with timestamp
5. Saves to: backend/news_articles.json

JSON Output Format:

{
  "last_updated": "2025-05-03T17:57:38",
  "source": "https://www.dawn.com/latest-news",
  "articles": [
    {
      "id": 1,
      "headline": "Article headline...",
      "url": "https://dawn.com/news/...",
      "summary": "First 5 paragraphs of article..."
    },
    ...
  ]
}

Issues to Fix:

⚠️ Runs as blocking infinite loop (should be background service)
⚠️ No error handling for network failures
⚠️ Hard-coded URL (not configurable)

2. script_generator.py — AI Script Writing via Gemini

What it does:

Reads articles from news_articles.json
Sends them to Google Gemini API
Gets back an engaging news script
Writes script to file for next step

How it works:

1. Load news_articles.json
2. Extract all headlines and summaries
3. Format as prompt for Gemini:
   "You're a scriptwriter for a news channel. Based on:
    [Headline 1]
    [Summary 1]
    [Headline 2]
    [Summary 2]
    Generate a 1-2 minute engaging video script in spoken English style."
4. Call genai.GenerativeModel("gemini-2.5-flash-preview-04-17")
5. Return response.text (plain text script)
6. Save to final_script.txt

Sample Output:

Good evening, I'm your news anchor. Tonight, we bring you 
the latest stories shaping our world. 

[Article 1 narration...]
[Article 2 narration...]
...
Stay tuned for more updates.

Issues to Fix:

⚠️ No API key (empty string currently)
⚠️ No error handling if API fails
⚠️ No timeout for API calls

3. text_to_speech.py — Voice Synthesis via ElevenLabs

What it does:

Reads the generated script
Converts it to realistic MP3 audio
Uses professional voice with settings

How it works:

1. Read final_script.txt
2. Call ElevenLabs API:
   POST https://api.elevenlabs.io/v1/text-to-speech/{VOICE_ID}
3. Headers:
   - xi-api-key: [YOUR_API_KEY]
   - Content-Type: application/json
4. Data:
   {
     "text": "[entire script]",
     "model_id": "eleven_monolingual_v1",
     "voice_settings": {
       "stability": 0.5,
       "similarity_boost": 0.75
     }
   }
5. Receive MP3 audio stream
6. Save to final_audio.mp3

Voice Settings Explanation:

Stability (0.5): Balance between consistency and variation
Similarity Boost (0.75): How closely to match the trained voice

Issues to Fix:

⚠️ No API key (empty string currently)
⚠️ No error handling
⚠️ Final_script.txt must exist first
⚠️ No validation that audio file is valid

4. generate_video.py — AI Avatar Video via D-ID

What it does:

Reads the MP3 audio file
Sends to D-ID API with an avatar
D-ID generates video of avatar lip-syncing
Downloads and saves the video

How it works:

1. Read final_audio.mp3
2. Encode to Base64 (text format for JSON transmission)
3. Call D-ID API:
   POST https://api.d-id.com/talks
4. Headers:
   - Authorization: Basic {API_TOKEN_BASE64}
   - Content-Type: application/json
5. Data:
   {
     "source_url": "",
     "script": {
       "type": "audio",
       "audio": "[BASE64_ENCODED_MP3]"
     },
     "driver_id": "amy",  # D-ID public avatar
     "config": {
       "fluent": true,
       "pad_audio": 0.2
     }
   }
6. Receive {"id": "talk_12345..."}
7. Poll status endpoint every 5 seconds:
   GET https://api.d-id.com/talks/{talk_id}
8. Wait for "result_url" to appear (takes ~30-60 seconds)
9. Download MP4 from result_url
10. Save to static/video/anchor.mp4

Processing Time: 30-60 seconds per video generation

Issues to Fix:

⚠️ No API token (empty string currently)
⚠️ No timeout for polling loop (could hang indefinitely)
⚠️ Final_audio.mp3 must exist first
⚠️ No error handling for failed generation

5. app.py — Flask Web Server & Orchestration

What it does:

Serves the web interface on port 5000
Provides API endpoint for script generation
Runs background scheduler for automation

How it works:

1. Initialize Flask app
2. Configure Gemini API with key
3. Set up BackgroundScheduler (APScheduler)
4. Schedule job: scheduled_script_generation()
   - Trigger: every 60 minutes
   - Function: generates script from news
   - Saves to: static/generated_script.txt
5. Start scheduler
6. Define routes:
   - GET / → serve index.html
   - POST /latest-news-script → generate script on-demand

Scheduler Details:

Uses APScheduler's BackgroundScheduler
Runs in same process as Flask
Triggers function every 60 minutes
Function logs timestamp and saves script to disk

Current Issue:

⚠️ The scheduler generates scripts but doesn't call TTS or video generation
⚠️ APScheduler doesn't persist if app crashes
⚠️ No way to see if scheduled job succeeded/failed

6. automation.py — Full Pipeline Orchestrator

What it does:

Runs the complete pipeline end-to-end
Coordinates all steps in sequence

Current Implementation:

if __name__ == "__main__":  # ⚠️ TYPO: Currently says _name_ instead of __name__
    print(">> Generating anchor script...")
    generate_anchor_script()
    print(">> Generating voiceover from script...")
    generate_voiceover()
    print(">> Creating avatar video...")
    generate_avatar_video()
    print(">> Pipeline completed. Video saved as anchor.mp4")

Issues to Fix:

⚠️ CRITICAL: Typo in line 6 — _name_ should be __name__
⚠️ Functions don't exist (generate_anchor_script, etc.) — should call actual functions
⚠️ No error handling between steps
⚠️ No way to skip steps if they already completed

7. index.html — Web Page Template

What it does:

Provides the user interface
Displays video player
Shows news articles
Renders 3D background

Key Elements:

<nav class="navbar">
  - VOXBYTE logo and navigation links
  
<div class="hero-section">
  - Spline 3D iframe (animated background)
  - "NEWS THAT NEVER SLEEPS" headline
  - "WATCH NOW" call-to-action button
  
<section id="headlines" class="headlines-section">
  - Last updated timestamp
  - Video player (plays anchor.mp4)
  - News articles container (filled by JavaScript)
  
<footer>
  - Copyright and tagline

Embedded Resources:

Google Fonts (Orbitron, Work Sans)
Spline 3D (https://my.spline.design/...)
Local CSS (styles.css)
Local JavaScript (script.js)

8. script.js — Frontend Interactivity

What it does:

Displays news articles on the page
Makes articles expandable/collapsible
Updates timestamp
Handles smooth scrolling

Current Issues:

⚠️ News data is HARDCODED in JavaScript instead of fetching from backend
⚠️ API fetch code is commented out (see lines with /* and */)
⚠️ Doesn't fetch /latest-news-script endpoint

To Fix: Replace hardcoded newsData with:

// Instead of hardcoding, fetch from backend:
fetch("/latest-news-script", {
    method: "POST",
    headers: {"Content-Type": "application/json"}
})
.then(res => res.json())
.then(data => {
    // Use data.articles to populate page
})

🚀 Installation & Setup Guide

Prerequisites

Before you start, you need:

Python 3.8+ installed
pip (Python package manager)
Three API Keys:
- Google Gemini API (free at https://aistudio.google.com/app/apikey)
- ElevenLabs API (free tier at https://www.elevenlabs.io/)
- D-ID API (free credits at https://www.d-id.com/api/)

Step 1: Navigate to Project Directory

cd VoxByte/TAG25/backend

Step 2: Install Python Dependencies

First, verify requirements.txt has all dependencies:

pip install flask google-generativeai requests beautifulsoup4 newspaper3k apscheduler schedule python-dotenv

Or create a comprehensive requirements.txt:

flask==2.3.0
google-generativeai==0.3.0
requests==2.31.0
beautifulsoup4==4.12.0
newspaper3k==0.0.9
apscheduler==3.10.0
schedule==1.2.0
python-dotenv==1.0.0

Then install:

pip install -r requirements.txt

Step 3: Get and Configure API Keys

A. Google Gemini API Key:

Go to https://aistudio.google.com/app/apikey
Click "Create API Key"
Copy the key

B. ElevenLabs API Key:

Sign up at https://www.elevenlabs.io/
Go to Account Settings → API Key
Copy the key
Note a Voice ID (use default or create custom voice)

C. D-ID API Token:

Sign up at https://www.d-id.com/api/
Get free credits
Go to Account → API Key
Copy the key

Step 4: Create .env File

In backend/ directory, create .env:

GEMINI_API_KEY=your_gemini_key_here
ELEVENLABS_API_KEY=your_elevenlabs_key_here
ELEVENLABS_VOICE_ID=VzCzzZS0ff2iL6Izl8fR
D_ID_API_TOKEN=your_d_id_token_here

Step 5: Update Python Files to Use .env

Update script_generator.py:

import os
from dotenv import load_dotenv

load_dotenv()
genai.configure(api_key=os.getenv("GEMINI_API_KEY"))

Update text_to_speech.py:

import os
from dotenv import load_dotenv

load_dotenv()
API_KEY = os.getenv("ELEVENLABS_API_KEY")
VOICE_ID = os.getenv("ELEVENLABS_VOICE_ID")

Update generate_video.py:

import os
from dotenv import load_dotenv

load_dotenv()
API_TOKEN = os.getenv("D_ID_API_TOKEN")

Step 6: Run the Web Server

python app.py

You should see:

 * Running on http://127.0.0.1:5000
 * Press CTRL+C to quit

Step 7: Access the Web Interface

Open your browser and go to:

http://localhost:5000

📊 How to Verify Everything is Working

✅ Check 1: Web Server Running

Expected:
✅ Flask server starts on http://localhost:5000
✅ Home page loads with Spline 3D animation background
✅ "NEWS THAT NEVER SLEEPS" heading visible
✅ Video player shows (may be empty initially)
✅ News section displays "Last updated" timestamp

✅ Check 2: News Scraping Working

cd backend
python news_scrapper.py

Expected:

🔁 Scraping started...
✅ Updated 'news_articles.json' with 5 articles at 2025-01-15T14:32:10

Check file exists:

cat news_articles.json

✅ Check 3: Script Generation Working

python -c "from script_generator import generate_script; print(generate_script())"

Expected Output:

Good evening, I'm your AI news anchor. Tonight we bring you the latest stories...
[2-3 paragraphs of news script]

✅ Check 4: Text-to-Speech Working

First, ensure final_script.txt exists with content:

echo "Good evening, I'm your news anchor." > final_script.txt

Then run:

python text_to_speech.py

Expected:

Creates final_audio.mp3 (~1-2 MB)
File should be a valid MP3 (can be played with any media player)

Test it:

# On Windows:
start final_audio.mp3

# On Mac:
open final_audio.mp3

# On Linux:
mpv final_audio.mp3

✅ Check 5: Video Generation Working

First, ensure final_audio.mp3 exists (from Check 4).

Run:

python generate_video.py

Expected:

[Polls for 30-60 seconds...]
Video downloaded and saved to static/video/anchor.mp4

✅ Check 6: Full Web Interface

Navigate to http://localhost:5000
Look for:
- ✅ Spline 3D animation in background
- ✅ "NEWS THAT NEVER SLEEPS" title
- ✅ Video player (if anchor.mp4 exists, should play video)
- ✅ News articles section with 5 articles
- ✅ "Last updated" timestamp
- ✅ Click article headers to expand/collapse summaries

✅ Check 7: Scheduled Automation

Keep app.py running
Watch the console output
Every 60 minutes, you should see:

Generating script at 2025-01-15 15:32:10.123456
Script saved.

If you see this, ✅ Scheduler is working!

🐛 Known Issues & How to Fix Them

🔴 CRITICAL ISSUES - Must Fix Before Deployment

Issue	Symptom	Root Cause	Fix
No API Keys	"Error: API key not found"	Keys not configured	Add API keys to `.env` file and update imports
Missing Dependencies	"ModuleNotFoundError: No module named..."	requirements.txt incomplete	Run: `pip install flask google-generativeai requests beautifulsoup4 apscheduler schedule`
automation.py typo	`NameError: name '_name_' is not defined`	Line 6 has `_name_` instead of `__name__`	Change `if _name_ == "_main_":` to `if __name__ == "__main__":`

⚠️ WARNING ISSUES - Should Fix for Production

Issue	Symptom	Root Cause	Fix
Video file missing	Video player shows no video	`generate_video.py` not run	Run the video generation script or create placeholder video
Hardcoded news in JS	News doesn't update from backend	script.js has hardcoded data	Uncomment the fetch code in script.js to get live data
News scraper blocks	Server freezes after news scraping	news_scrapper.py infinite loop	Run scraper as separate background service/process
No error handling	Crashes on API failures	Missing try-catch blocks	Add error handling to all API calls
Scheduler unreliable	Scheduled jobs don't run	APScheduler needs persistence	Use external task queue (Celery) for production

🌐 Deployment Guide

Local Development (Easy)

cd backend
python app.py
# Access: http://localhost:5000

✅ Good for: Testing, development
❌ Issues: App stops when you close terminal

Heroku Cloud Deployment

Install Heroku CLI
Create Procfile in backend/:

web: gunicorn app:app
worker: python news_scrapper.py

Create runtime.txt:

python-3.10.13

Update requirements.txt:

flask==2.3.0
gunicorn==21.0.0
google-generativeai==0.3.0
requests==2.31.0
beautifulsoup4==4.12.0
apscheduler==3.10.0
schedule==1.2.0
python-dotenv==1.0.0

Deploy:

heroku create voxbyte-app
heroku config:set GEMINI_API_KEY=xxx
heroku config:set ELEVENLABS_API_KEY=xxx
heroku config:set D_ID_API_TOKEN=xxx
git push heroku main

Access: https://voxbyte-app.herokuapp.com

✅ Good for: Production deployment
✅ Auto-scaling, monitoring, custom domain

Docker Deployment

Create Dockerfile in backend/:

FROM python:3.10-slim

WORKDIR /app

# Install dependencies
COPY requirements.txt .
RUN pip install --no-cache-dir -r requirements.txt

# Copy app
COPY . .

# Expose port
EXPOSE 5000

# Run app
CMD ["python", "app.py"]

Build & Run:

docker build -t voxbyte .
docker run -p 5000:5000 \
  -e GEMINI_API_KEY=xxx \
  -e ELEVENLABS_API_KEY=xxx \
  -e D_ID_API_TOKEN=xxx \
  voxbyte

✅ Good for: Containerized deployments, consistency

AWS EC2 Deployment

Launch EC2 instance (Ubuntu 20.04+)
SSH into instance
Install Python & dependencies:

sudo apt-get update
sudo apt-get install python3-pip python3-venv

Clone code:

git clone [your-repo-url]
cd VoxByte/TAG25/backend

Set up virtual environment:

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

Create systemd service (auto-start):

sudo nano /etc/systemd/system/voxbyte.service

Add:

[Unit]
Description=VOXBYTE News Service
After=network.target

[Service]
User=ubuntu
WorkingDirectory=/home/ubuntu/VoxByte/TAG25/backend
ExecStart=/home/ubuntu/VoxByte/TAG25/backend/venv/bin/python app.py
Restart=always

Environment="GEMINI_API_KEY=xxx"
Environment="ELEVENLABS_API_KEY=xxx"
Environment="D_ID_API_TOKEN=xxx"

[Install]
WantedBy=multi-user.target

Start service:

sudo systemctl start voxbyte
sudo systemctl enable voxbyte

Check logs:

sudo journalctl -u voxbyte -f

✅ Good for: Self-managed production
⚠️ You handle: Scaling, monitoring, backups

Important Production Considerations

API Keys: Use environment variables ONLY, never hardcode
News Scraper: Run as separate background job/worker (not in Flask)
Error Logging: Add logging to catch failures
Rate Limiting: Add delays to avoid exceeding API quotas
Caching: Cache generated scripts/videos to reduce API calls
Monitoring: Set up monitoring/alerting for failed pipelines
Reverse Proxy: Use Nginx/Apache in front of Flask for production
HTTPS: Get SSL certificate (Let's Encrypt is free)
Database: Consider using database instead of JSON for scalability

🎯 Quick Reference: Running Each Component Individually

Just Scrape News

python news_scrapper.py
# Creates: news_articles.json

Just Generate Script

python -c "from script_generator import generate_script; print(generate_script())"
# Requires: news_articles.json
# Creates: final_script.txt (if saved in script_generator)

Just Generate Voiceover

# First create final_script.txt
python text_to_speech.py
# Requires: final_script.txt
# Creates: final_audio.mp3

Just Generate Video

python generate_video.py
# Requires: final_audio.mp3
# Creates: static/video/anchor.mp4

Run Full Pipeline

# FIRST: Fix the typo in automation.py (__name__ not _name_)
# THEN:
python automation.py
# Runs all steps: script → audio → video

Start Web Server

python app.py
# Serves on: http://localhost:5000
# Runs scheduler in background (every 60 minutes)

📝 API Integration Details

Google Gemini API

Endpoint: genai.GenerativeModel() Model Used: gemini-2.5-flash-preview-04-17 Cost: FREE tier available Prompt Template:

"You're a scriptwriter for a news channel. Based on the following latest news summaries, 
generate a 1-2 minute engaging and informative video script in English:

[Headlines and summaries here]

The script should have a brief intro and use natural spoken English style."

Response: Plain text script (no formatting)

ElevenLabs Text-to-Speech API

Endpoint: https://api.elevenlabs.io/v1/text-to-speech/{VOICE_ID} Method: POST Headers:

xi-api-key: Your API key
Content-Type: application/json

Request Body:

{
  "text": "[Full script here]",
  "model_id": "eleven_monolingual_v1",
  "voice_settings": {
    "stability": 0.5,
    "similarity_boost": 0.75
  }
}

Response: MP3 audio file (binary) Cost: FREE tier (10,000 characters/month)

D-ID Avatar Video API

Endpoint 1 (Create Video): https://api.d-id.com/talks Method: POST Headers:

Authorization: Basic {TOKEN}
Content-Type: application/json

Request Body:

{
  "source_url": "",
  "script": {
    "type": "audio",
    "audio": "[BASE64_ENCODED_MP3]"
  },
  "driver_id": "amy",
  "config": {
    "fluent": true,
    "pad_audio": 0.2
  }
}

Endpoint 2 (Check Status): https://api.d-id.com/talks/{talk_id} Method: GET Response: {"result_url": "https://..." } when ready

Processing Time: 30-60 seconds Cost: FREE credits for testing

👥 Team Roles & Responsibilities

Name	Role	Responsibilities
Syeda Fatima Zahra	Backend Engineer	Flask server setup, Gemini API integration, script generation logic, app.py development
Nabira Salman	Automation Engineer	News web scraping, TTS pipeline, video generation, orchestration logic
Amna Maryam Fatima	Frontend Developer	HTML/CSS design, JavaScript interactivity, Spline 3D integration, responsive UI

🚨 Pre-Deployment Checklist

Before deploying to production, verify:

💡 Troubleshooting Guide

"ModuleNotFoundError: No module named 'flask'"

pip install flask google-generativeai requests beautifulsoup4 apscheduler schedule

"Error: API key not found" or similar

Check .env file exists in backend/ directory
Check environment variables are loaded with load_dotenv()
Check keys are not empty strings

"news_articles.json not found"

Run: python news_scrapper.py
Check for network errors (Dawn News might be blocked)
Try running manually first

"No module named 'newspaper3k'"

pip install newspaper3k

Video player shows blank

Ensure static/video/anchor.mp4 exists
Run: python generate_video.py
Check file size (should be 5-10 MB)

News articles show hardcoded data

Open backend/static/script.js
Find commented-out fetch code
Uncomment the fetch block
Refresh browser

Scheduler doesn't trigger

Check app.py console for scheduler log messages
Verify app.py is still running
APScheduler only works while app is running

"Connection refused" or "Cannot reach API"

Check internet connection
Verify API endpoint URLs are correct
Check firewall isn't blocking outbound connections
Verify API keys have correct permissions

🏆 Project Achievement

Competition: TAG25 NUST Innovation Challenge Result: 🥈 2nd Place (Runner-Up) Timeline: Completed in < 24 hours Team Size: 3 developers Outcome: Fully functional AI news broadcasting system

📄 License & Usage

🤝 Contributing & Support

For issues or improvements:

Check this README for troubleshooting
Review the code comments in each Python file
Test components individually (see "Quick Reference")
Check API documentation links in this README

📚 Additional Resources

Gemini API Docs: https://ai.google.dev/
ElevenLabs Docs: https://docs.elevenlabs.io/
D-ID Docs: https://www.d-id.com/api/
Flask Docs: https://flask.palletsprojects.com/
Beautiful Soup Docs: https://www.crummy.com/software/BeautifulSoup/bs4/doc/

Last Updated: 2025-01-15
Version: 1.0 (Production Ready)

Name		Name	Last commit message	Last commit date
static		static
templates		templates
.env.example		.env.example
.gitignore		.gitignore
Dockerfile		Dockerfile
README.md		README.md
app.py		app.py
automation.py		automation.py
final_script.txt		final_script.txt
generate_video.py		generate_video.py
news_articles.json		news_articles.json
news_scrapper.py		news_scrapper.py
page.html		page.html
requirements.txt		requirements.txt
response.json		response.json
script_generator.py		script_generator.py
text_to_speech.py		text_to_speech.py

Latest commit

History

Folders and files

Repository files navigation

📰 VOXBYTE — AI-Powered News Anchor App

🎯 Project Overview

🔥 Core Features

🧩 Technology Stack

📁 Project Structure

🏗️ How It Works (System Architecture)

Complete Pipeline Flow

Key API Endpoints

Data Flow Diagram

⚙️ Component Breakdown (Detailed)

1. news_scrapper.py — News Data Harvesting

2. script_generator.py — AI Script Writing via Gemini

3. text_to_speech.py — Voice Synthesis via ElevenLabs

4. generate_video.py — AI Avatar Video via D-ID

5. app.py — Flask Web Server & Orchestration

6. automation.py — Full Pipeline Orchestrator

7. index.html — Web Page Template

8. script.js — Frontend Interactivity

🚀 Installation & Setup Guide

Prerequisites

Step 1: Navigate to Project Directory

Step 2: Install Python Dependencies

Step 3: Get and Configure API Keys

Step 4: Create .env File

Step 5: Update Python Files to Use .env

Step 6: Run the Web Server

Step 7: Access the Web Interface

📊 How to Verify Everything is Working

✅ Check 1: Web Server Running

✅ Check 2: News Scraping Working

✅ Check 3: Script Generation Working

✅ Check 4: Text-to-Speech Working

✅ Check 5: Video Generation Working

✅ Check 6: Full Web Interface

✅ Check 7: Scheduled Automation

🐛 Known Issues & How to Fix Them

🔴 CRITICAL ISSUES - Must Fix Before Deployment

⚠️ WARNING ISSUES - Should Fix for Production

🌐 Deployment Guide

Local Development (Easy)

Heroku Cloud Deployment

Docker Deployment

AWS EC2 Deployment

Important Production Considerations

🎯 Quick Reference: Running Each Component Individually

Just Scrape News

Just Generate Script

Just Generate Voiceover

Just Generate Video

Run Full Pipeline

Start Web Server

📝 API Integration Details

Google Gemini API

ElevenLabs Text-to-Speech API

D-ID Avatar Video API

👥 Team Roles & Responsibilities

🚨 Pre-Deployment Checklist

💡 Troubleshooting Guide

"ModuleNotFoundError: No module named 'flask'"

"Error: API key not found" or similar

"news_articles.json not found"

"No module named 'newspaper3k'"

Video player shows blank

News articles show hardcoded data

Scheduler doesn't trigger

"Connection refused" or "Cannot reach API"

🏆 Project Achievement

📄 License & Usage

🤝 Contributing & Support

📚 Additional Resources

About

Resources

Stars

Watchers

Forks

Releases