YouTube Captions Fixer

yt-captions-fixer.userscript.js

// ==UserScript==
// @name         YouTube Captions Fix
// @namespace    http://tampermonkey.net/
// @version      2025-03-13
// @description  Blank out all instances of YouTube's censorship/sound detection in auto-generated captions
// @author       luna (h-lunah)
// @match        https://www.youtube.com/*
// @match        https://www.youtube-nocookie.com/*
// @match        https://m.youtube.com/*
// @match        https://music.youtube.com/*
// @icon         https://cdn.discordapp.com/emojis/1236984687126384640.png?v=2
// @grant        none
// @run-at       document-start
// ==/UserScript==

(function() {
    'use strict';
    /* match all instances of YouTube's censorship or sound detection
     * example captions given substitute "fuck" (censored) for 1 or more spaces */
    const censorRe = new RegExp(/\[(.*?)\]/gm);        // censorship signatures, [Music], [Applause], [Laughter], etc. for any language
    const spaceRe = new RegExp(/\s{2,}/gm);            // spaces leftover after signature cleanup, eg. "well fuck you man"
    const punctRe = new RegExp(/\s(?=[!?.])/gm);       // spaces leftover before punctuation marks preceded by signature markers
    const altRe = new RegExp(/^foreign$|^thank you$|^laughs$/gm); // alternatives for [Music] and [Laughter]

    const removeCaptions = () => {
      // interface with any caption element, even preview ones
      var captionBox = document.querySelector(".captions-text");
      var segmentBox = document.querySelector(".ytd-transcript-segment-renderer");
      var captions = document.getElementsByClassName("ytp-caption-segment");
      var transcriptCaptions = document.getElementsByClassName("segment-text");

      // match detected captions
      for ( const caption of captions ) {
          
        caption.style.whiteSpace = "normal";

        // blank out detected signatures
        if ( caption.innerText.search(censorRe) !== -1 ) {
            console.log(`removed: ${caption.innerText.match(censorRe).join(", ")}`);
            caption.innerText = caption.innerText.replaceAll(censorRe, "");
        }

        // blank out other signatures for sound detection
        if ( caption.innerText.search(altRe) !== -1 ) {
            console.log(`removed alternative signature: ${caption.innerText.match(altRe)}`);
            caption.innerText = caption.innerText.replaceAll(altRe, "");
        }

        // clean up leftovers after signature blanking
        if ( caption.innerText.search(spaceRe) !== -1 ) {
            console.log("removed double whitespace");
            caption.innerText = caption.innerText.replaceAll(spaceRe, " ");
        }

        // clean up excess whitespace after [censors] followed by punctuation (human captions)
        if ( caption.innerText.search(punctRe) !== -1 ) {
            console.log("removed whitespace before punctutation mark");
            caption.innerText = caption.innerText.replaceAll(punctRe, "");
        }

        if ( caption.innerText.search(/Click for settings/gm) !== -1 ) {
            caption.innerText = caption.innerText.replaceAll(/Click for settings/gm, "These captions have been cleaned up");
        }

        /* remove fully blanked out caption lines
         * applies to caption lines during non-spoken section of videos,  eg. https://www.youtube.com/watch?v=1HJRGygc1yc&t=50s
         *
         * "fuck fuck fucking football"
         * "[Music]                   " */
        if ( !caption.innerText ) {
          caption.parentElement.remove();
          console.log("removed stray blank caption");
        }

        /* remove fully blanked captions
         * applies to music-only videos, eg. https://www.youtube.com/watch?v=OpaKHVOEDiE
         *
         * "[Music]" */
        if ( captionBox && !captionBox.childNodes.length ) {
          captionBox.parentElement.remove();
          console.log("removed blank caption box");
        }
      }

      // match caption transcripts
      for ( const transcriptCaption of transcriptCaptions ) {
        if ( transcriptCaption.innerText.search(censorRe) !== -1 ) {
            console.log(`removed: ${transcriptCaption.innerText.match(censorRe).join(", ")}`);
            transcriptCaption.innerText = transcriptCaption.innerText.replaceAll(censorRe, "");
        }

        // blank out other signatures for sound detection
        if ( transcriptCaption.innerText.search(altRe) !== -1 ) {
            console.log(`removed alternative signature: ${transcriptCaption.innerText.match(altRe)}`);
            transcriptCaption.innerText = transcriptCaption.innerText.replaceAll(altRe, "");
        }

        // clean up leftovers after signature blanking
        if ( transcriptCaption.innerText.search(spaceRe) !== -1 ) {
            console.log("removed double whitespace");
            transcriptCaption.innerText = transcriptCaption.innerText.replaceAll(spaceRe, " ");
        }

        // clean up excess whitespace after [censors] followed by punctuation (human captions)
        if ( transcriptCaption.innerText.search(punctRe) !== -1 ) {
            console.log("removed whitespace before punctutation mark");
            transcriptCaption.innerText = transcriptCaption.innerText.replaceAll(punctRe, "");
        }

        /* remove fully blanked out transcript lines
         * applies to sections where there was only music/unknown sounds
         * "[Music]" */
        if ( !transcriptCaption.innerText ) {
            transcriptCaption.parentNode.remove();
            console.log("removed blank transcript box");
        }
      }
    }

    // Optimized removals are done using MutationObservers. You will not see any markers flash at all.
    const setupObserver = () => {
      // Make sure the <body> exists
      if (document.body) {
        const observer = new MutationObserver(() => removeCaptions());

        // Observe the <body>
        observer.observe(document.body, {
          childList: true,
          subtree: true
        });

        // Run filter on what's present, there isn't anything loaded at this stage though.
        removeCaptions();
      } else {
        // Wait for <body> to appear
        setTimeout(setupObserver, 10);
      }
    };

    // Enable caption cleaning
    setupObserver();
})();