@storyteller-platform/align 0.1.56 → 0.1.58

package/dist/align/ctc/Aligner.cjs

@@ -0,0 +1,800 @@
+"use strict";
+var __create = Object.create;
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+var __getProtoOf = Object.getPrototypeOf;
+var __hasOwnProp = Object.prototype.hasOwnProperty;
+var __export = (target, all) => {
+  for (var name in all)
+    __defProp(target, name, { get: all[name], enumerable: true });
+};
+var __copyProps = (to, from, except, desc) => {
+  if (from && typeof from === "object" || typeof from === "function") {
+    for (let key of __getOwnPropNames(from))
+      if (!__hasOwnProp.call(to, key) && key !== except)
+        __defProp(to, key, { get: () => from[key], enumerable: !(desc = __getOwnPropDesc(from, key)) || desc.enumerable });
+  }
+  return to;
+};
+var __toESM = (mod, isNodeMode, target) => (target = mod != null ? __create(__getProtoOf(mod)) : {}, __copyProps(
+  // If the importer is in node compatibility mode or this is not an ESM
+  // file that has been converted to a CommonJS file using a Babel-
+  // compatible transform (i.e. "__esModule" has not been set), then set
+  // "default" to the CommonJS "module.exports" for node compatibility.
+  isNodeMode || !mod || !mod.__esModule ? __defProp(target, "default", { value: mod, enumerable: true }) : target,
+  mod
+));
+var __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var Aligner_exports = {};
+__export(Aligner_exports, {
+  Aligner: () => Aligner
+});
+module.exports = __toCommonJS(Aligner_exports);
+var import_promises = require("node:fs/promises");
+var import_posix = require("node:path/posix");
+var import_itertools = require("itertools");
+var import_memoize = __toESM(require("memoize"), 1);
+var import_epub = require("@storyteller-platform/epub");
+var import_ghost_story = require("@storyteller-platform/ghost-story");
+var import_media_types = require("@storyteller-platform/media-types");
+var import_parseDom = require("../../markup/parseDom.cjs");
+var import_segmentation = require("../../markup/segmentation.cjs");
+var import_transform = require("../../markup/transform.cjs");
+var import_getSentenceRanges = require("../getSentenceRanges.cjs");
+var import_interpolateSentenceRanges = require("../interpolateSentenceRanges.cjs");
+var import_slugify = require("../slugify.cjs");
+var import_emissions = require("./emissions.cjs");
+var import_forcedAlign = require("./forcedAlign.cjs");
+var import_mediaOverlay = require("./mediaOverlay.cjs");
+var import_report = require("./report.cjs");
+var import_search = require("./search.cjs");
+const MAX_SEARCH_LENGTH = 4e4;
+class Aligner {
+  constructor(epub, audiofiles, emissionsReader, granularity, textRef, languageOverride, logger) {
+    this.epub = epub;
+    this.audiofiles = audiofiles;
+    this.emissionsReader = emissionsReader;
+    this.languageOverride = languageOverride;
+    this.logger = logger;
+    this.audiofileFrames = /* @__PURE__ */ new Map();
+    this.audiofileEmissionTimings = /* @__PURE__ */ new Map();
+    for (const [i, audiofile] of (0, import_itertools.enumerate)(audiofiles)) {
+      const [, framesStart, framesEnd] = emissionsReader.frameMap[i];
+      const timings = emissionsReader.timings[i];
+      const length = framesEnd - framesStart;
+      this.audiofileFrames.set(audiofile, [framesStart, framesEnd]);
+      this.audioFileDurations[audiofile] = (0, import_emissions.framesToSeconds)(length, timings);
+      this.audiofileEmissionTimings.set(audiofile, timings);
+    }
+    this.getChapterSentences = (0, import_memoize.default)(this.getChapterSentences.bind(this));
+    this.granularity = granularity ?? "sentence";
+    this.textRef = textRef ?? "id-fragment";
+  }
+  audiofileFrames;
+  audiofileEmissionTimings;
+  totalDuration = 0;
+  alignedChapters = [];
+  timing = (0, import_ghost_story.createAggregator)();
+  granularity;
+  textRef;
+  audioFileDurations = {};
+  report = {
+    chapters: [],
+    unalignedChapters: [],
+    audioFiles: [],
+    unalignedAudioFiles: []
+  };
+  async getChapterSentences(chapterId) {
+    const chapterXml = await this.epub.readXhtmlItemContents(chapterId);
+    const original = (0, import_parseDom.parseDom)(import_epub.Epub.getXhtmlBody(chapterXml));
+    const inlined = (0, import_transform.inlineFootnotes)(original);
+    const lifted = (0, import_transform.liftText)(inlined.root);
+    const segmentation = await (0, import_segmentation.segmentChapter)(lifted.result, {
+      primaryLocale: this.languageOverride ?? await this.epub.getLanguage()
+    });
+    return segmentation.filter((s) => s.text.match(/\S/));
+  }
+  async writeAlignedChapter(alignedChapter) {
+    const locale = this.languageOverride ?? await this.epub.getLanguage() ?? new Intl.Locale("en-US");
+    const { chapter, sentenceRanges, wordRanges, xml } = alignedChapter;
+    const sentences = await this.getChapterSentences(chapter.id);
+    const {
+      sentenceIdToBlockFragment,
+      sentenceIdToFragment,
+      wordIdToFragment,
+      wordRangeMap
+    } = (0, import_mediaOverlay.buildFragmentMaps)(
+      sentenceRanges,
+      wordRanges,
+      sentences,
+      locale,
+      this.textRef,
+      this.granularity
+    );
+    await this.addChapterAudio(sentenceRanges);
+    const mediaOverlayId = await this.addMediaOverlay(
+      chapter,
+      sentenceRanges,
+      wordRangeMap,
+      sentenceIdToBlockFragment,
+      sentenceIdToFragment,
+      wordIdToFragment
+    );
+    await this.epub.updateManifestItem(chapter.id, {
+      ...chapter,
+      mediaOverlay: mediaOverlayId
+    });
+    await this.epub.writeXhtmlItemContents(chapter.id, xml);
+    const chapterDuration = (0, import_getSentenceRanges.getChapterDuration)(sentenceRanges);
+    await this.epub.addMetadata({
+      type: "meta",
+      properties: {
+        property: "media:duration",
+        refines: `#${mediaOverlayId}`
+      },
+      value: import_epub.Epub.formatSmilDuration(chapterDuration)
+    });
+  }
+  async addChapterAudio(sentenceRanges) {
+    const audiofiles = Array.from(
+      new Set(sentenceRanges.map(({ audiofile }) => audiofile))
+    );
+    await Promise.all(
+      audiofiles.map(async (audiofile) => {
+        const { name, base } = (0, import_posix.parse)(audiofile);
+        const id = `audio_${name}`;
+        const manifest = await this.epub.getManifest();
+        if (id in manifest) return;
+        const epubAudioFilename = `Audio/${base}`;
+        const duration = this.audioFileDurations[audiofile];
+        this.totalDuration += duration;
+        const audio = await (0, import_promises.readFile)(audiofile);
+        const mediaType = import_media_types.MediaType.fromPath(base)?.mime;
+        await this.epub.addManifestItem(
+          {
+            id,
+            href: epubAudioFilename,
+            mediaType
+          },
+          audio
+        );
+      })
+    );
+  }
+  async addMediaOverlay(chapter, sentenceRanges, wordRangeMap, sentenceIdToBlockFragment, sentenceIdToFragment, wordIdToFragment) {
+    const { name: chapterStem } = (0, import_posix.parse)(chapter.href);
+    const mediaOverlayId = `${chapter.id}_overlay`;
+    await this.epub.addManifestItem(
+      {
+        id: mediaOverlayId,
+        href: `MediaOverlays/${chapterStem}.smil`,
+        mediaType: "application/smil+xml"
+      },
+      (0, import_mediaOverlay.createMediaOverlay)(
+        chapter,
+        this.granularity,
+        sentenceRanges,
+        wordRangeMap,
+        sentenceIdToBlockFragment,
+        sentenceIdToFragment,
+        wordIdToFragment
+      ),
+      "xml"
+    );
+    return mediaOverlayId;
+  }
+  addChapterReport(spineIndex, chapter, chunk, score, chapterSentences, sentenceCount, sentenceRanges, startSentence, endSentence, alignedEmissionsStart, alignedEmissionsEnd) {
+    const audioFiles = (0, import_report.buildAudioFileContexts)(sentenceRanges);
+    this.report.chapters.push({
+      href: chapter.href,
+      spineIndex,
+      chunk,
+      score,
+      transcriptionOffset: alignedEmissionsStart,
+      endTranscriptionOffset: alignedEmissionsEnd,
+      firstMatchedSentenceId: startSentence,
+      firstMatchedSentenceContext: {
+        prevSentence: chapterSentences[startSentence - 1]?.text ?? null,
+        // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+        matchedSentence: chapterSentences[startSentence].text,
+        nextSentence: chapterSentences[startSentence + 1]?.text ?? null
+      },
+      lastMatchedSentenceId: endSentence,
+      lastMatchedSentenceContext: {
+        prevSentence: chapterSentences[endSentence - 1]?.text ?? null,
+        // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+        matchedSentence: chapterSentences[endSentence].text,
+        nextSentence: chapterSentences[endSentence + 1]?.text ?? null
+      },
+      chapterSentenceCount: sentenceCount,
+      alignedSentenceCount: sentenceRanges.length,
+      audioFiles
+    });
+    (0, import_report.updateAudioFileReports)(this.report, audioFiles, this.audioFileDurations);
+  }
+  async getSentenceRanges(chapterId, slugifiedChapter, sentenceStarts, wordStarts, sentenceIndexOffset, emissionsStart, emissionsEnd, anchors, paddedChapterBoundaries, unpaddedChapterBoundaries) {
+    const SENTENCE_PAD = 25;
+    const WORD_PAD = 8;
+    const windowAnchors = anchors.filter(
+      (a) => a.position > emissionsStart && a.position < emissionsEnd && a.position - emissionsStart >= a.offset && emissionsEnd - a.position >= slugifiedChapter.length - a.offset
+    ).map((a) => ({ ...a, position: a.position - emissionsStart }));
+    const alignments = (0, import_forcedAlign.ctcForcedAlign)(
+      slugifiedChapter,
+      await this.emissionsReader.readFrames(emissionsStart, emissionsEnd),
+      windowAnchors,
+      this.emissionsReader.vocab,
+      this.emissionsReader.blankId,
+      (labelOffset, alignments2) => sentenceStarts.filter(
+        ([s, e]) => s >= labelOffset && e < labelOffset + alignments2.length && (0, import_forcedAlign.detectDroppedLabels)(alignments2, s - labelOffset, e - labelOffset)
+      )
+    );
+    const finiteScores = alignments.filter(({ score: score2 }) => score2 !== -Infinity);
+    const unscaledScore = finiteScores.reduce((acc, { score: score2 }) => acc + score2, 0) / finiteScores.length;
+    const score = finiteScores.length > 0 ? Math.max((unscaledScore + 10) * 10, 0) : 0;
+    const {
+      sentenceRanges,
+      skippedSentences,
+      firstFoundSentence,
+      lastFoundSentence,
+      firstConfirmedSentence,
+      lastConfirmedSentence
+    } = this.buildSentenceRanges(
+      alignments,
+      sentenceStarts,
+      sentenceIndexOffset,
+      emissionsStart,
+      SENTENCE_PAD,
+      chapterId,
+      paddedChapterBoundaries,
+      unpaddedChapterBoundaries
+    );
+    const wordRanges = this.buildWordRanges(
+      alignments,
+      wordStarts,
+      sentenceIndexOffset,
+      emissionsStart,
+      WORD_PAD,
+      skippedSentences,
+      chapterId
+    );
+    if (firstFoundSentence === null || firstConfirmedSentence === null) {
+      return {
+        score,
+        sentenceRanges: [],
+        wordRanges: [],
+        alignedEmissionsStart: emissionsStart,
+        alignedEmissionsEnd: emissionsStart,
+        firstFoundSentence: sentenceIndexOffset,
+        lastFoundSentence: sentenceIndexOffset
+      };
+    }
+    return {
+      score,
+      sentenceRanges,
+      wordRanges,
+      alignedEmissionsStart: alignments[sentenceStarts[firstConfirmedSentence][0]].startFrame + emissionsStart,
+      alignedEmissionsEnd: alignments[sentenceStarts[lastConfirmedSentence][1]].endFrame + emissionsStart,
+      firstFoundSentence: sentenceIndexOffset + firstFoundSentence,
+      lastFoundSentence: sentenceIndexOffset + lastFoundSentence
+    };
+  }
+  buildSentenceRanges(alignments, sentenceStarts, sentenceIndexOffset, emissionsStart, padFrames, chapterId, paddedChapterBoundaries, unpaddedChapterBoundaries) {
+    const sentenceRanges = [];
+    const skippedSentences = /* @__PURE__ */ new Set();
+    let firstFoundSentence = null;
+    let lastFoundSentence = 0;
+    let lastFoundSentenceEnd = 0;
+    let firstConfirmedSentence = null;
+    let lastConfirmedSentence = 0;
+    for (const [i, [startOffset, endOffset]] of (0, import_itertools.enumerate)(sentenceStarts)) {
+      if (endOffset < startOffset) {
+        skippedSentences.add(i);
+        continue;
+      }
+      if (endOffset < paddedChapterBoundaries[0] || startOffset >= paddedChapterBoundaries[1]) {
+        skippedSentences.add(i);
+        continue;
+      }
+      const goodChars = alignments.slice(startOffset, endOffset + 1).reduce((acc, a) => acc + (a.score > -4 ? 1 : 0), 0);
+      if (goodChars / (endOffset + 1 - startOffset) < 0.5) {
+        skippedSentences.add(i);
+        continue;
+      }
+      const sentenceLength = endOffset + 1 - startOffset;
+      const framesPerChar = (alignments[endOffset].endFrame - alignments[startOffset].startFrame) / sentenceLength;
+      if (sentenceLength > 8 && (framesPerChar < 2 || framesPerChar > 15)) {
+        skippedSentences.add(i);
+        continue;
+      }
+      const startFrame = this.refineStartFrame(
+        alignments[startOffset].startFrame,
+        lastFoundSentenceEnd,
+        padFrames,
+        emissionsStart
+      );
+      firstFoundSentence ??= i;
+      lastFoundSentence = i;
+      lastFoundSentenceEnd = alignments[endOffset].endFrame;
+      const { position: start, audiofile } = this.frameToAudioPosition(
+        startFrame,
+        emissionsStart
+      );
+      const { position: end } = this.frameToAudioPosition(
+        alignments[endOffset].endFrame,
+        emissionsStart
+      );
+      sentenceRanges.push({
+        id: sentenceIndexOffset + i,
+        chapterId,
+        start,
+        end,
+        audiofile
+      });
+      if (startOffset <= unpaddedChapterBoundaries[1] && endOffset >= unpaddedChapterBoundaries[0]) {
+        firstConfirmedSentence ??= i;
+        lastConfirmedSentence = i;
+      }
+    }
+    return {
+      sentenceRanges,
+      skippedSentences,
+      firstFoundSentence,
+      lastFoundSentence,
+      lastFoundSentenceEnd,
+      firstConfirmedSentence,
+      lastConfirmedSentence
+    };
+  }
+  buildWordRanges(alignments, wordStarts, sentenceIndexOffset, emissionsStart, padFrames, skippedSentences, chapterId) {
+    const wordRanges = [];
+    let lastFoundWordEnd = 0;
+    for (const [i, ws] of (0, import_itertools.enumerate)(wordStarts)) {
+      if (skippedSentences.has(i)) continue;
+      const ranges = [];
+      for (const [j, [startOffset, endOffset]] of (0, import_itertools.enumerate)(ws)) {
+        if (endOffset < startOffset) continue;
+        const goodChars = alignments.slice(startOffset, endOffset + 1).reduce((acc, a) => acc + (a.score > -4 ? 1 : 0), 0);
+        if (goodChars / (endOffset + 1 - startOffset) < 0.5) {
+          continue;
+        }
+        const startFrame = this.refineStartFrame(
+          alignments[startOffset].startFrame,
+          lastFoundWordEnd,
+          padFrames,
+          emissionsStart
+        );
+        lastFoundWordEnd = alignments[endOffset].endFrame;
+        const { position: start, audiofile } = this.frameToAudioPosition(
+          startFrame,
+          emissionsStart
+        );
+        const { position: end } = this.frameToAudioPosition(
+          alignments[endOffset].endFrame,
+          emissionsStart
+        );
+        ranges.push({
+          id: j,
+          sentenceId: sentenceIndexOffset + i,
+          chapterId,
+          start,
+          end,
+          audiofile
+        });
+      }
+      if (ranges.length) wordRanges.push(ranges);
+    }
+    return wordRanges;
+  }
+  frameToAudioPosition(frame, emissionsStart) {
+    const globalFrame = frame + emissionsStart;
+    for (const [audiofile, [fileStart, fileEnd]] of this.audiofileFrames) {
+      if (globalFrame < fileEnd) {
+        return {
+          position: (0, import_emissions.framesToSeconds)(
+            globalFrame - fileStart,
+            // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+            this.audiofileEmissionTimings.get(audiofile)
+          ),
+          audiofile
+        };
+      }
+    }
+    throw new Error(`Frame ${globalFrame} is beyond the end of the emissions`);
+  }
+  refineStartFrame(startFrame, prevEndFrame, padFrames, emissionsStart) {
+    const { audiofile } = this.frameToAudioPosition(startFrame, emissionsStart);
+    const [start] = this.audiofileFrames.get(audiofile);
+    const windowFloor = start - emissionsStart;
+    return Math.max(prevEndFrame + 1, startFrame - padFrames, windowFloor);
+  }
+  async alignChapter(spineIndex, chapterId, chunk, sentenceIndexOffset, chapterSentences, slugifiedChapter, sentenceStarts, wordStarts, emissionsStart, emissionsEnd, anchors, paddedChapterBoundaries, unpaddedChapterBoundaries) {
+    const timing = (0, import_ghost_story.createTiming)();
+    timing.start("read contents");
+    const manifest = await this.epub.getManifest();
+    const chapter = manifest[chapterId];
+    if (!chapter)
+      throw new Error(
+        `Failed to align chapter: could not find chapter with id ${chapterId} in manifest`
+      );
+    const chapterXml = await this.epub.readXhtmlItemContents(chapterId);
+    timing.start("align sentences");
+    const {
+      score,
+      sentenceRanges,
+      wordRanges,
+      alignedEmissionsStart,
+      alignedEmissionsEnd,
+      firstFoundSentence,
+      lastFoundSentence
+    } = await this.getSentenceRanges(
+      chapterId,
+      slugifiedChapter,
+      sentenceStarts,
+      wordStarts,
+      sentenceIndexOffset,
+      emissionsStart,
+      emissionsEnd,
+      anchors,
+      paddedChapterBoundaries,
+      unpaddedChapterBoundaries
+    );
+    timing.end("align sentences");
+    const storytellerStylesheetUrl = (0, import_posix.relative)(
+      (0, import_posix.dirname)(chapter.href),
+      "Styles/storyteller-readaloud.css"
+    );
+    import_epub.Epub.addLinkToXhtmlHead(chapterXml, {
+      rel: "stylesheet",
+      href: storytellerStylesheetUrl,
+      type: "text/css"
+    });
+    this.alignedChapters.push({
+      chapter,
+      chunk,
+      xml: chapterXml,
+      sentenceRanges,
+      wordRanges,
+      startOffset: alignedEmissionsStart,
+      endOffset: alignedEmissionsEnd
+    });
+    this.addChapterReport(
+      spineIndex,
+      chapter,
+      chunk,
+      score,
+      chapterSentences,
+      sentenceStarts.length,
+      sentenceRanges,
+      firstFoundSentence,
+      lastFoundSentence,
+      alignedEmissionsStart,
+      alignedEmissionsEnd
+    );
+    return {
+      lastSentenceRange: sentenceRanges.at(-1) ?? null,
+      alignedEmissionsEnd,
+      timing
+    };
+  }
+  getMatchedBoundaries() {
+    return this.alignedChapters.toSorted((a, b) => a.startOffset - b.startOffset).map(({ startOffset, endOffset }) => [startOffset, endOffset]);
+  }
+  narrowToAvailableBoundary(boundary) {
+    const available = [
+      -1,
+      ...this.getMatchedBoundaries().flat(),
+      Infinity
+    ];
+    const withinBoundary = [];
+    for (let i = 0; i < available.length - 1; i += 2) {
+      const [start, end] = [available[i], available[i + 1]];
+      if (boundary.start <= start && boundary.end >= start || boundary.start <= end && boundary.end >= end) {
+        withinBoundary.push([
+          Math.max(boundary.start, start + 1),
+          Math.min(boundary.end, end - 1)
+        ]);
+      }
+    }
+    const largestBoundary = (0, import_itertools.max)(withinBoundary, ([start, end]) => end - start);
+    if (!largestBoundary) return { start: boundary.start, end: boundary.end };
+    return { start: largestBoundary[0], end: largestBoundary[1] };
+  }
+  async alignBook(onProgress) {
+    const { locale } = await this.prepareAlignment();
+    await this.processSpine(onProgress, locale);
+    await this.finalizeAlignment();
+    return this.timing;
+  }
+  async prepareAlignment() {
+    const locale = this.languageOverride ?? await this.epub.getLanguage() ?? new Intl.Locale("en-US");
+    this.timing.setMetadata("language", locale.toString());
+    this.timing.setMetadata("granularity", this.granularity);
+    return { locale };
+  }
+  async processSpine(onProgress, locale) {
+    const spine = await this.epub.getSpineItems();
+    const manifest = await this.epub.getManifest();
+    for (let index = 0; index < spine.length; index++) {
+      onProgress?.(index / spine.length);
+      const spineItem = spine[index];
+      this.logger?.info(
+        `Aligning epub item #${index} : ${(0, import_posix.basename)(spineItem.href)}`
+      );
+      const results = await this.processSpineItem(
+        index,
+        spineItem,
+        manifest,
+        locale
+      );
+      results.forEach((result) => this.timing.add(result.timing.summary()));
+    }
+  }
+  async processSpineItem(index, spineItem, manifest, locale) {
+    const chapterId = spineItem.id;
+    if (manifest[chapterId]?.properties?.includes("nav")) {
+      this.report.unalignedChapters.push({
+        href: spineItem.href,
+        reason: "is-nav"
+      });
+      return [];
+    }
+    const chapterSentences = await this.getChapterSentences(chapterId);
+    const slugifiedChapterSentences = [];
+    const chapterMappings = [];
+    let slugifiedChapterLength = 0;
+    for (const chapterSentence of chapterSentences) {
+      const { result, mapping } = await (0, import_slugify.slugify)(
+        chapterSentence.text,
+        locale,
+        ""
+      );
+      slugifiedChapterSentences.push(result);
+      slugifiedChapterLength += result.length;
+      chapterMappings.push(mapping);
+    }
+    if (chapterSentences.length === 0 || slugifiedChapterLength === 0) {
+      this.logger?.info(`Chapter #${index} has no text; skipping`);
+      this.report.unalignedChapters.push({
+        href: spineItem.href,
+        reason: "no-text"
+      });
+      return [];
+    }
+    if (chapterSentences.length < 2 && // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+    chapterSentences[0].words.length < 4) {
+      this.logger?.info(`Chapter #${index} is fewer than four words; skipping`);
+      this.report.unalignedChapters.push({
+        href: spineItem.href,
+        reason: "too-short"
+      });
+      return [];
+    }
+    const boundaries = await (0, import_search.findCtcBoundaries)(
+      slugifiedChapterSentences.join(""),
+      this.emissionsReader,
+      this.getMatchedBoundaries()
+    );
+    const evidenceRange = boundaries?.evidenceRange;
+    const matchedRange = boundaries?.matchedRange;
+    const chunkCount = Math.ceil(slugifiedChapterLength / MAX_SEARCH_LENGTH);
+    const chunkLength = Math.ceil(slugifiedChapterLength / chunkCount);
+    const chunks = [
+      { firstSentence: 0, sentences: [] }
+    ];
+    let c = 0;
+    for (const [i, s] of (0, import_itertools.enumerate)(slugifiedChapterSentences)) {
+      if (c >= chunkLength * chunks.length && // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+      chunks.at(-1).sentences.length > 0) {
+        chunks.push({ firstSentence: i, sentences: [] });
+      }
+      chunks.at(-1).sentences.push(s);
+      c += s.length;
+    }
+    if (chunks.length > 1) {
+      this.logger?.info(
+        `Chapter #${index} is ${slugifiedChapterLength} chars; aligning in ${chunks.length} chunks`
+      );
+    }
+    const results = [];
+    const firstAlignedChapterIndex = this.alignedChapters.length;
+    for (const [
+      chunkIndex,
+      { firstSentence, sentences: chunkSentences }
+    ] of (0, import_itertools.enumerate)(chunks)) {
+      let pos = 0;
+      const slugifiedChapterRanges = [];
+      for (const sent of chunkSentences) {
+        slugifiedChapterRanges.push([pos, pos + sent.length - 1]);
+        pos += sent.length;
+      }
+      const slugifiedWordStarts = chunkSentences.map((_, i) => {
+        const sent = chapterSentences[firstSentence + i];
+        const cursor = chapterMappings[firstSentence + i].cursor();
+        return sent.words.entries.filter((word) => !word.isPunctuation).map((word) => [
+          cursor.map(word.startOffset - sent.charRange.start) + // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+          slugifiedChapterRanges[i][0],
+          cursor.map(word.endOffset - 1 - sent.charRange.start) + // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+          slugifiedChapterRanges[i][0]
+        ]);
+      });
+      const chunkString = chunkSentences.join("");
+      const boundaries2 = await (0, import_search.findCtcBoundaries)(
+        chunkString,
+        this.emissionsReader,
+        this.getMatchedBoundaries()
+      );
+      if (!boundaries2) {
+        this.logger?.info(
+          `Could not find chapter #${index} in the transcripton`
+        );
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "not-found",
+          spineIndex: index,
+          score: 0,
+          chapterSentenceCount: chunkSentences.length,
+          start: chapterSentences.slice(firstSentence, firstSentence + 3).map((s) => s.text).join("").slice(0, 80),
+          end: chapterSentences.slice(
+            Math.max(
+              firstSentence,
+              firstSentence + chunkSentences.length - 3
+            ),
+            firstSentence + chunkSentences.length
+          ).map((s) => s.text).join("").slice(-80)
+        });
+        continue;
+      }
+      const { start, end } = this.narrowToAvailableBoundary(boundaries2);
+      if (start === end) {
+        this.logger?.info(
+          `Could not find chapter #${index} in the transcripton`
+        );
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "not-found",
+          spineIndex: index,
+          score: 0,
+          chapterSentenceCount: chunkSentences.length,
+          start: chapterSentences.slice(firstSentence, firstSentence + 3).map((s) => s.text).join("").slice(0, 80),
+          end: chapterSentences.slice(
+            Math.max(
+              firstSentence,
+              firstSentence + chunkSentences.length - 3
+            ),
+            firstSentence + chunkSentences.length
+          ).map((s) => s.text).join("").slice(-80)
+        });
+        continue;
+      }
+      const lastChunkSlugStart = slugifiedChapterLength - chunkString.length;
+      const chunkEvidenceHead = chunkIndex === 0 ? evidenceRange?.[0] ?? boundaries2.evidenceRange[0] : boundaries2.evidenceRange[0];
+      const chunkEvidenceTail = chunkIndex === chunks.length - 1 && evidenceRange ? evidenceRange[1] - lastChunkSlugStart : boundaries2.evidenceRange[1];
+      const chunkMatchedHead = chunkIndex === 0 ? matchedRange?.[0] ?? boundaries2.matchedRange[0] : boundaries2.matchedRange[0];
+      const chunkMatchedTail = chunkIndex === chunks.length - 1 && matchedRange ? matchedRange[1] - lastChunkSlugStart : boundaries2.matchedRange[1];
+      const result = await this.alignChapter(
+        index,
+        chapterId,
+        chunkIndex,
+        firstSentence,
+        chapterSentences,
+        chunkString,
+        slugifiedChapterRanges,
+        slugifiedWordStarts,
+        start,
+        end,
+        boundaries2.anchors,
+        [chunkEvidenceHead, chunkEvidenceTail],
+        [chunkMatchedHead, chunkMatchedTail]
+      );
+      results.push(result);
+    }
+    const chunkEntries = this.alignedChapters.slice(firstAlignedChapterIndex);
+    if (chunkEntries.length > 1) {
+      const claimed = chunkEntries.filter((e) => e.sentenceRanges.length > 0);
+      const offsetSources = claimed.length > 0 ? claimed : chunkEntries;
+      const first = chunkEntries[0];
+      this.alignedChapters.splice(
+        firstAlignedChapterIndex,
+        chunkEntries.length,
+        {
+          chapter: first.chapter,
+          chunk: 0,
+          xml: first.xml,
+          sentenceRanges: chunkEntries.flatMap((e) => e.sentenceRanges),
+          wordRanges: chunkEntries.flatMap((e) => e.wordRanges),
+          startOffset: Math.min(...offsetSources.map((e) => e.startOffset)),
+          endOffset: Math.max(...offsetSources.map((e) => e.endOffset))
+        }
+      );
+    }
+    return results;
+  }
+  async finalizeAlignment() {
+    const audioOrderedChapters = this.alignedChapters.toSorted((a, b) => {
+      const firstRangeA = a.sentenceRanges[0];
+      const firstRangeB = b.sentenceRanges[0];
+      if (!firstRangeA) return 1;
+      if (!firstRangeB) return -1;
+      const firstAudiofileIndexA = this.audiofiles.indexOf(
+        firstRangeA.audiofile
+      );
+      const firstAudiofileIndexB = this.audiofiles.indexOf(
+        firstRangeB.audiofile
+      );
+      if (firstAudiofileIndexA === firstAudiofileIndexB) {
+        return firstRangeA.start - firstRangeB.start;
+      }
+      return firstAudiofileIndexA - firstAudiofileIndexB;
+    });
+    const sentenceRanges = [];
+    const chapterSentenceCounts = {};
+    const sentenceLengths = {};
+    for (const alignedChapter of audioOrderedChapters) {
+      sentenceRanges.push(...alignedChapter.sentenceRanges);
+      const sentences = await this.getChapterSentences(
+        alignedChapter.chapter.id
+      );
+      chapterSentenceCounts[alignedChapter.chapter.id] = sentences.length;
+      for (const [id, sentence] of (0, import_itertools.enumerate)(sentences)) {
+        sentenceLengths[(0, import_interpolateSentenceRanges.slotKey)({ chapterId: alignedChapter.chapter.id, id })] = sentence.text.length;
+      }
+    }
+    const interpolated = (0, import_interpolateSentenceRanges.interpolateSentenceRanges)(
+      sentenceRanges,
+      chapterSentenceCounts,
+      this.audioFileDurations,
+      sentenceLengths
+    );
+    const expanded = (0, import_getSentenceRanges.expandEmptySentenceRanges)(interpolated);
+    const collapsed = await (0, import_getSentenceRanges.collapseSentenceRangeGaps)(expanded);
+    let collapsedStart = 0;
+    for (const alignedChapter of audioOrderedChapters) {
+      const sentences = await this.getChapterSentences(
+        alignedChapter.chapter.id
+      );
+      const finalSentenceRanges = collapsed.slice(
+        collapsedStart,
+        collapsedStart + sentences.length
+      );
+      alignedChapter.sentenceRanges = finalSentenceRanges;
+      for (const [i, wordRanges] of (0, import_itertools.enumerate)(alignedChapter.wordRanges)) {
+        alignedChapter.wordRanges[i] = (0, import_getSentenceRanges.expandEmptySentenceRanges)(wordRanges);
+      }
+      await this.writeAlignedChapter(alignedChapter);
+      collapsedStart += sentences.length;
+    }
+    for (const audiofile of this.audiofiles) {
+      if (!this.report.audioFiles.some(({ filepath }) => filepath === audiofile)) {
+        this.report.unalignedAudioFiles.push({ filepath: audiofile });
+      }
+    }
+    await this.epub.addMetadata({
+      type: "meta",
+      properties: { property: "media:duration" },
+      value: import_epub.Epub.formatSmilDuration(this.totalDuration)
+    });
+    await this.epub.addMetadata({
+      type: "meta",
+      properties: { property: "media:active-class" },
+      value: "-epub-media-overlay-active"
+    });
+    await this.epub.addManifestItem(
+      {
+        id: "storyteller_readaloud_styles",
+        href: "Styles/storyteller-readaloud.css",
+        mediaType: "text/css"
+      },
+      `
+.-epub-media-overlay-active {
+  background-color: #ffb;
+}
+      `,
+      "utf-8"
+    );
+  }
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  Aligner
+});