Converts WEBVTT subtitles (vtt) to plain text.

webvtt2txt.py

#!/usr/bin/env python3

"""
Converts WEBVTT subtitles (vtt) to plain text.

It removes all time related info as well as duplicated and empty lines.
"""
# Author: Artur Martins <[email protected]>
# Version: 1.0
# Date: 2023-Oct-25

import re
import argparse
import os

ENCODING_TYPE = "utf-8"
HEADER = "WEBVTT"


def clean_line(line: str) -> str:
    """
    Remove all WebVTT tags and time codes from the given line.

    Args:
        line (str): The line of text to clean.

    Returns:
        str: The cleaned line with all tags and time codes removed and leading/trailing whitespace stripped.
    """
    # Remove all WebVTT tags and time codes
    cleaned_line = re.sub(r"<.*?>", "", line)
    cleaned_line = re.sub(r"\d{2}:\d{2}:\d{2}\.\d{3}", "", cleaned_line)
    return cleaned_line.strip()


def convert_webvtt_to_text(input_path: str, output_path: str, verbose: bool) -> None:
    """
    Convert a WebVTT file to plain text.

    Args:
        input_path (str): The path to the WebVTT input file.
        output_path (str): The path to the output text file.
        verbose (bool): If True, print the cleaned lines as they are written.

    Returns:
        None
    """
    last_written_line = ""
    with open(input_path, "r", encoding=ENCODING_TYPE) as infile, open(
        output_path, "w", encoding=ENCODING_TYPE
    ) as outfile:
        lines = infile.readlines()
        for line in lines:
            line = line.strip()
            # Skip time lines or WebVTT header
            if "-->" in line or line == HEADER:
                continue
            # Skip empty lines
            if not line:
                continue
            cleaned_line = clean_line(line)
            if cleaned_line and cleaned_line != last_written_line:
                if verbose:
                    print(f"Writing: {cleaned_line}")
                outfile.write(cleaned_line + "\n")
                last_written_line = cleaned_line


if __name__ == "__main__":
    parser = argparse.ArgumentParser(description="Convert WebVTT to text.")
    parser.add_argument(
        "-i", "--input", required=True, help="Path to the input WebVTT file."
    )
    parser.add_argument(
        "-o",
        "--output",
        help="Path to the output text file. Defaults to input file name with .txt extension.",
    )

    parser.add_argument(
        "-v", "--verbose", action="store_true", help="Enable verbose output."
    )

    args = parser.parse_args()

    if args.output is None:
        base_name = os.path.splitext(args.input)[0]
        args.output = f"{base_name}.txt"

    if args.verbose:
        print(f"Converting {args.input} to {args.output}...")

    convert_webvtt_to_text(args.input, args.output, args.verbose)

Nice code - can I reuse this in a package?

What would be proper attribution?

Sorry @epogrebnyak, I completely missed your message.

Sure, no problem. Feel free to leave my name and email in the script as is. Good luck!

Thank you for this.

I wanted to use it as a filter so that I could:

yt-dlp --skip-download --sub-langs en --convert-subs vtt -o ~/.tmp.subtitle <video_url> | vtt2txt; cat ~/.tmp.subtitle.en.vtt | vtt2txt | fabric -ps extract_wisdom

So I made this into this filter

You're welcome, @mholtzhausen ! Your filter it's pretty cool! I like it! I will use it for myself as well :)