Python script to convert Youtube subtitles to txt files

convert_subtitle.py

#!/usr/bin/env python3
import json
import re

#
# Convert Youtube subtitle to txt file
#
# Look for `https://www.youtube.com/api/timedtext` in your browser network tab
# on a youtube video. Copy the response into a subtitle.json file and save this
# script in the same folder. 
#
# run `$ python3 convert_subtitle.py` and a txt file should be generated with your
# subtitle!
# 

def parse_subtitles_to_transcript(json_file, output_file):
    """
    Parse subtitle JSON file and create a clean transcript.
    """
    try:
        with open(json_file, 'r', encoding='utf-8') as f:
            data = json.load(f)
        
        # Extract all text segments from events
        transcript_parts = []
        
        for event in data.get('events', []):
            # Skip events without segments
            if 'segs' not in event:
                continue
            
            # Extract text from each segment
            segment_text = ""
            for seg in event['segs']:
                if 'utf8' in seg:
                    text = seg['utf8']
                    # Skip standalone newlines and music annotations
                    if text == '\n':
                        continue
                    if text.startswith('[') and text.endswith(']'):
                        continue
                    segment_text += text
            
            # Add the segment text if it's not empty
            if segment_text.strip():
                transcript_parts.append(segment_text.strip())
        
        # Join all parts into a coherent transcript
        transcript = ' '.join(transcript_parts)
        
        # Clean up the transcript
        # Remove extra spaces
        transcript = re.sub(r'\s+', ' ', transcript)
        
        # Add proper punctuation and paragraph breaks for readability
        # Split into sentences based on context and add proper formatting
        sentences = []
        current_sentence = ""
        
        words = transcript.split()
        for i, word in enumerate(words):
            current_sentence += word + " "
            
            # Simple sentence boundary detection
            # This is basic but should work for most cases
            if (len(current_sentence.split()) > 15 or 
                (len(current_sentence.split()) > 8 and i < len(words) - 1 and 
                 words[i+1][0].isupper())):
                sentences.append(current_sentence.strip())
                current_sentence = ""
        
        # Add the last sentence if any
        if current_sentence.strip():
            sentences.append(current_sentence.strip())
        
        # Format the transcript with proper line breaks
        formatted_transcript = ""
        for i, sentence in enumerate(sentences):
            formatted_transcript += sentence
            if not sentence.endswith('.'):
                formatted_transcript += "."
            formatted_transcript += "\n\n"
        
        # Write to output file
        with open(output_file, 'w', encoding='utf-8') as f:
            f.write(formatted_transcript.strip())
        
        print(f"Successfully converted subtitles to transcript: {output_file}")
        print(f"Total sentences: {len(sentences)}")
        
    except Exception as e:
        print(f"Error converting subtitles: {e}")

if __name__ == "__main__":
    parse_subtitles_to_transcript('subtitle.json', 'transcript.txt')