server-plugins/generate-captions.js

import { fileURLToPath } from 'url';
import { readFileSync, createWriteStream, existsSync } from 'fs';
import { Plugin } from './plugin.js';
import path from 'path';
import fs from 'fs';

export class CaptionPlugin extends Plugin {
  constructor(name, options) {
    super(name, options);
  }
  async applyPrerender(originalManuscript, jobId) {
    let transcript = originalManuscript.transcript
    for (let item of transcript) {
      let audioCaptionFileFileName = path.basename(item.audioCaptionFile)
      if (path.extname(audioCaptionFileFileName) == '.ass') {
        continue
      }
      let originalCaption = path.join(process.cwd(), item.audioCaptionFile)
      if (!fs.existsSync(originalCaption)) {
        originalCaption = path.join(process.cwd(), 'public', audioCaptionFileFileName)
      }
      if (!originalCaption) continue;
      let outputCaptionFile = originalCaption.replace('.json', '.ass')
      await this.generateCaptions(
        {
          ...this.options,
          captionFilePath: originalCaption,
          outputFilePath: outputCaptionFile,
        }
      )
      item._audioCaptionFile = item.audioCaptionFile
      item.audioCaptionFile = item.audioCaptionFile.replace('.json', '.ass')
    }
  }
  async applyPostrender(originalManuscript, jobId, outFiles) {
  }



  /**

   * Generate ASS subtitle file with word highlighting

   * @param {Object} options

   * @param {string} options.captionFilePath - Path to input JSON caption file

   * @param {string} options.outputFilePath - Path to output ASS file

   * @param {string} options.capitalize - capitalize the font. upper, full-upper, small, none

   * @param {number} [options.tiltDegrees=8] - Tilt angle in degrees (alternates between +/-)

   * @param {number} [options.translateY=200] - Distance from bottom in pixels

   * @param {number} [options.widthPercent=80] - Width percentage for text centering (0-100)

   * @param {string} [options.fontName='Impact'] - Font name

   * @param {number} [options.fontSize=72] - Font size

   * @param {number} [options.wordsPerGroup=4] - Number of words per caption group

   * @param {number} [options.videoWidth=1920] - Video width for positioning

   * @param {number} [options.videoHeight=1080] - Video height for positioning

   * @returns {Promise<string>} Path to generated ASS file

   */
  async generateCaptions(options) {
    const {
      captionFilePath,
      outputFilePath,
      tiltDegrees = 8,
      translateY = 200,
      widthPercent = 80,
      fontName = 'Impact',
      fontSize = 72,
      capitalize = 'upper',
      wordsPerGroup = 4,
      videoWidth = 1920,
      videoHeight = 1080,
      fontColor = '#FFFFFF',
      fontHighlightColor = '#00FF00'
    } = options;

    const assFontColor = hexToASSColor(fontColor);
    const assHighlightColor = hexToASSColor(fontHighlightColor);
    const assHighlightColorInline = `${assHighlightColor}&`;
    const assFontColorInline = `${assFontColor}&`;

    // Read and parse JSON file
    const jsonData = JSON.parse(readFileSync(captionFilePath, 'utf-8'));
    const transcript = jsonData.transcript || '';
    let words = jsonData.words || [];

    if (words.length === 0) {
      throw new Error('No words found in caption file');
    }

    if (capitalize == 'full-upper') {
      words = words.map(w => ({ ...w, word: w.word.toUpperCase() }));
    }
    else if (capitalize == 'upper') {
      words = words.map(w => ({ ...w, word: w.word.charAt(0).toUpperCase() + w.word.slice(1) }));
    }
    else if (capitalize == 'small') {
      words = words.map(w => ({ ...w, word: w.word.toLowerCase() }));
    }
    // Assign sentence indices to words
    words = assignSentenceToWords(words, transcript);

    // Calculate margins for centering within width percentage
    const totalMargin = videoWidth * (1 - widthPercent / 100);
    const sideMargin = Math.floor(totalMargin / 2);

    // Create output stream
    const output = createWriteStream(outputFilePath);

    // Write header with calculated margins
    output.write(
      createASSHeader(
        videoWidth,
        videoHeight,
        fontName,
        fontSize,
        translateY,
        sideMargin,
        sideMargin,
        assFontColor,
        assHighlightColor
      )
    );

    // Process words in groups respecting sentence boundaries
    let i = 0;
    let groupIdx = 0;

    while (i < words.length) {
      const currentSentence = words[i].sentence_idx || 0;

      // Collect words for this group (up to wordsPerGroup, same sentence only)
      const wordGroup = [];
      let j = i;

      while (j < words.length && wordGroup.length < wordsPerGroup) {
        if ((words[j].sentence_idx || 0) === currentSentence) {
          wordGroup.push(words[j]);
          j++;
        } else {
          break; // Stop at sentence boundary
        }
      }

      if (wordGroup.length === 0) {
        i++;
        continue;
      }

      // Alternate tilt
      const currentTilt = groupIdx % 2 === 0 ? tiltDegrees : -tiltDegrees;
      const tiltTag = `{\\frz${currentTilt}}`;

      // Calculate positioning for centering
      const posTag = sideMargin > 0 ? `{\\an2\\pos(${videoWidth / 2},${videoHeight - translateY})}` : '';

      // Get the full group duration (from first word start to last word end)
      const groupStart = wordGroup[0].start;
      const groupEnd = wordGroup[wordGroup.length - 1].end;

      // For each word in the group, create an event with highlighting
      // Use the FULL GROUP duration for each event to ensure no gaps
      for (let wordIdx = 0; wordIdx < wordGroup.length; wordIdx++) {
        const wordObj = wordGroup[wordIdx];
        const wordStart = wordObj.start;
        const wordEnd = wordIdx < wordGroup.length - 1 ? wordGroup[wordIdx + 1].start : wordObj.end;

        // Build the caption text with highlighting
        const captionParts = wordGroup.map((w, idx) => {
          if (idx === wordIdx) {
            // Current word - highlighted in green
            return `{\\c${assHighlightColorInline}}${w.word}{\\c${assFontColorInline}}`;
          } else {
            // Other words - white
            return w.word;
          }
        });

        const captionText = tiltTag + posTag + captionParts.join(' ');

        // Write dialogue line with timing from current word start to next word start (or group end)
        // This ensures continuous display with no gaps between words
        output.write(`Dialogue: 0,${formatTimestampASS(wordStart)},${formatTimestampASS(wordEnd)},Default,,0,0,0,,${captionText}\n`);
      }

      i = j;
      groupIdx++;
    }

    output.end();

    return new Promise((resolve, reject) => {
      output.on('finish', () => {
        this.log(`Generated ${path.basename(outputFilePath)} captions`);
        resolve(outputFilePath);
      });
      output.on('error', reject);
    });
  }
}


/**

 * Format seconds to ASS timestamp format (H:MM:SS.cc)

 * @param {number} seconds 

 * @returns {string}

 */
function formatTimestampASS(seconds) {
  const hours = Math.floor(seconds / 3600);
  const minutes = Math.floor((seconds % 3600) / 60);
  const secs = seconds % 60;
  return `${hours}:${minutes.toString().padStart(2, '0')}:${secs.toFixed(2).padStart(5, '0')}`;
}

/**

 * Split transcript into sentences

 * @param {string} transcript 

 * @returns {string[]}

 */
function splitIntoSentences(transcript) {
  const parts = transcript.split(/([.!?]+)\s+/);
  const result = [];

  for (let i = 0; i < parts.length - 1; i += 2) {
    if (i + 1 < parts.length) {
      result.push(parts[i] + parts[i + 1]);
    } else {
      result.push(parts[i]);
    }
  }

  if (parts.length % 2 === 1) {
    result.push(parts[parts.length - 1]);
  }

  return result;
}

/**

 * Assign sentence index to each word

 * @param {Array} words 

 * @param {string} transcript 

 * @returns {Array}

 */
function assignSentenceToWords(words, transcript) {
  const sentences = splitIntoSentences(transcript);

  let wordIdx = 0;
  sentences.forEach((sentence, sentIdx) => {
    const sentenceWords = sentence.split(/\s+/);
    sentenceWords.forEach(() => {
      if (wordIdx < words.length) {
        words[wordIdx].sentence_idx = sentIdx;
        wordIdx++;
      }
    });
  });

  return words;
}

/**

 * Create ASS file header with styles

 * @param {number} videoWidth 

 * @param {number} videoHeight 

 * @param {string} fontName 

 * @param {number} fontSize 

 * @param {number} marginV 

 * @returns {string}

 */
function createASSHeader(

  videoWidth = 1920,

  videoHeight = 1080,

  fontName = 'Impact',

  fontSize = 72,

  marginV = 200,

  marginL = 10,

  marginR = 10,

  primaryColor = '&H00FFFFFF',

  highlightColor = '&H0000FF00'

) {
  return `[Script Info]

Title: Word-by-Word Captions

ScriptType: v4.00+

WrapStyle: 0

PlayResX: ${videoWidth}

PlayResY: ${videoHeight}

ScaledBorderAndShadow: yes



[V4+ Styles]

Format: Name, Fontname, Fontsize, PrimaryColour, SecondaryColour, OutlineColour, BackColour, Bold, Italic, Underline, StrikeOut, ScaleX, ScaleY, Spacing, Angle, BorderStyle, Outline, Shadow, Alignment, MarginL, MarginR, MarginV, Encoding

Style: Default,${fontName},${fontSize},${primaryColor},&H000000FF,&H00000000,&H80000000,-1,0,0,0,100,100,0,0,1,3,2,2,${marginL},${marginR},${marginV},1

Style: Highlight,${fontName},${fontSize},${highlightColor},&H000000FF,&H00000000,&H80000000,-1,0,0,0,100,100,0,0,1,3,2,2,${marginL},${marginR},${marginV},1



[Events]

Format: Layer, Start, End, Style, Name, MarginL, MarginR, MarginV, Effect, Text

`;
}

function hexToASSColor(hexValue) {
  if (typeof hexValue !== 'string') {
    throw new Error('fontColor values must be hex strings like #RRGGBB');
  }

  const normalized = hexValue.trim().replace('#', '');
  if (!/^[0-9a-fA-F]{6}$/.test(normalized)) {
    throw new Error(`Invalid hex color provided: ${hexValue}`);
  }

  const r = normalized.slice(0, 2);
  const g = normalized.slice(2, 4);
  const b = normalized.slice(4, 6);
  return `&H00${b}${g}${r}`.toUpperCase();
}