[Examples] Enhance real-time transcription with VAD, word timestamps, and CLI options

[vasanthrpjan1-boop]

This PR builds upon #2696 and significantly enhances the real-time transcription example with production-ready features.

New Features

Voice Activity Detection (VAD)

• Only transcribes when speech is detected, saving compute resources
• Configurable energy threshold (--energy-threshold)
• Automatic speech segmentation based on silence detection

Word-Level Timestamps

• --word-timestamps flag shows timing for each word
• Useful for subtitling and precise audio alignment

Speaker Change Detection (Experimental)

• --detect-speakers provides hints when speaker changes are detected
• Based on pause pattern analysis

Audio Device Selection

• --list-devices to show available microphones
• --device-id to select a specific input device

Enhanced User Experience

• Live audio level visualization with color-coded bar
• Beautiful terminal UI with box-drawn headers
• Duplicate transcript filtering using similarity scoring
• Transcript saving with optional timestamps (--output, --timestamps)

Usage Examples

Basic usage

python examples/real_time_transcription.py

With word timestamps

python examples/real_time_transcription.py --word-timestamps

Save transcript with timestamps

python examples/real_time_transcription.py --output notes.txt --timestamps

Use specific model and language

python examples/real_time_transcription.py --model small --language es## Changes

• examples/real_time_transcription.py - Complete rewrite with new features
• README.md - Updated documentation with usage examples