@storyteller-platform/align 0.1.24 → 0.1.26

package/dist/align/align.cjs

@@ -87,8 +87,11 @@ var import_audiobook = require("@storyteller-platform/audiobook");
 var import_epub = require("@storyteller-platform/epub");
 var import_ghost_story = require("@storyteller-platform/ghost-story");
 var import_ffmpeg = require("../common/ffmpeg.cjs");
+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_search = require("./search.cjs");
 var import_slugify = require("./slugify.cjs");
 var import_textFragments = require("./textFragments.cjs");
@@ -172,12 +175,12 @@ class Aligner {
   };
   async getChapterSentences(chapterId) {
     const chapterXml = await this.epub.readXhtmlItemContents(chapterId);
-    const { result: segmentation } = await (0, import_segmentation.getXhtmlSegmentation)(
-      import_epub.Epub.getXhtmlBody(chapterXml),
-      {
-        primaryLocale: this.languageOverride ?? await this.epub.getLanguage()
-      }
-    );
+    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) {
@@ -505,16 +508,25 @@ class Aligner {
     });
     const sentenceRanges = [];
     const chapterSentenceCounts = {};
+    const audioFileDurations = {};
     for (const alignedChapter of audioOrderedChapters) {
       sentenceRanges.push(...alignedChapter.sentenceRanges);
+      for (const sentenceRange of sentenceRanges) {
+        if (!(sentenceRange.audiofile in audioFileDurations)) {
+          audioFileDurations[sentenceRange.audiofile] = await (0, import_ffmpeg.getTrackDuration)(
+            sentenceRange.audiofile
+          );
+        }
+      }
       const sentences = await this.getChapterSentences(
         alignedChapter.chapter.id
       );
       chapterSentenceCounts[alignedChapter.chapter.id] = sentences.length;
     }
-    const interpolated = await (0, import_getSentenceRanges.interpolateSentenceRanges)(
+    const interpolated = (0, import_interpolateSentenceRanges.interpolateSentenceRanges)(
       sentenceRanges,
-      chapterSentenceCounts
+      chapterSentenceCounts,
+      audioFileDurations
     );
     const expanded = (0, import_getSentenceRanges.expandEmptySentenceRanges)(interpolated);
     const collapsed = await (0, import_getSentenceRanges.collapseSentenceRangeGaps)(expanded);
@@ -525,7 +537,7 @@ class Aligner {
       );
       const finalSentenceRanges = collapsed.slice(
         collapsedStart,
-        collapsedStart + sentences.length - 1
+        collapsedStart + sentences.length
       );
       alignedChapter.sentenceRanges = finalSentenceRanges;
       for (const [i, wordRanges] of (0, import_itertools.enumerate)(alignedChapter.wordRanges)) {

package/dist/align/align.js

@@ -16,15 +16,17 @@ import {
   createTiming
 } from "@storyteller-platform/ghost-story";
 import { getTrackDuration } from "../common/ffmpeg.js";
-import { getXhtmlSegmentation } from "../markup/segmentation.js";
+import { parseDom } from "../markup/parseDom.js";
+import { segmentChapter } from "../markup/segmentation.js";
+import { inlineFootnotes, liftText } from "../markup/transform.js";
 import {
   collapseSentenceRangeGaps,
   expandEmptySentenceRanges,
   getChapterDuration,
   getSentenceRanges,
-  interpolateSentenceRanges,
   mapTranscriptionTimeline
 } from "./getSentenceRanges.js";
+import { interpolateSentenceRanges } from "./interpolateSentenceRanges.js";
 import { findBoundaries } from "./search.js";
 import { slugify } from "./slugify.js";
 import { TextFragmentTrie } from "./textFragments.js";
@@ -108,12 +110,12 @@ class Aligner {
   };
   async getChapterSentences(chapterId) {
     const chapterXml = await this.epub.readXhtmlItemContents(chapterId);
-    const { result: segmentation } = await getXhtmlSegmentation(
-      Epub.getXhtmlBody(chapterXml),
-      {
-        primaryLocale: this.languageOverride ?? await this.epub.getLanguage()
-      }
-    );
+    const original = parseDom(Epub.getXhtmlBody(chapterXml));
+    const inlined = inlineFootnotes(original);
+    const lifted = liftText(inlined.root);
+    const segmentation = await segmentChapter(lifted.result, {
+      primaryLocale: this.languageOverride ?? await this.epub.getLanguage()
+    });
     return segmentation.filter((s) => s.text.match(/\S/));
   }
   async writeAlignedChapter(alignedChapter) {
@@ -441,16 +443,25 @@ class Aligner {
     });
     const sentenceRanges = [];
     const chapterSentenceCounts = {};
+    const audioFileDurations = {};
     for (const alignedChapter of audioOrderedChapters) {
       sentenceRanges.push(...alignedChapter.sentenceRanges);
+      for (const sentenceRange of sentenceRanges) {
+        if (!(sentenceRange.audiofile in audioFileDurations)) {
+          audioFileDurations[sentenceRange.audiofile] = await getTrackDuration(
+            sentenceRange.audiofile
+          );
+        }
+      }
       const sentences = await this.getChapterSentences(
         alignedChapter.chapter.id
       );
       chapterSentenceCounts[alignedChapter.chapter.id] = sentences.length;
     }
-    const interpolated = await interpolateSentenceRanges(
+    const interpolated = interpolateSentenceRanges(
       sentenceRanges,
-      chapterSentenceCounts
+      chapterSentenceCounts,
+      audioFileDurations
     );
     const expanded = expandEmptySentenceRanges(interpolated);
     const collapsed = await collapseSentenceRangeGaps(expanded);
@@ -461,7 +472,7 @@ class Aligner {
       );
       const finalSentenceRanges = collapsed.slice(
         collapsedStart,
-        collapsedStart + sentences.length - 1
+        collapsedStart + sentences.length
       );
       alignedChapter.sentenceRanges = finalSentenceRanges;
       for (const [i, wordRanges] of enumerate(alignedChapter.wordRanges)) {

package/dist/align/getSentenceRanges.cjs

@@ -23,7 +23,6 @@ __export(getSentenceRanges_exports, {
   findEndTimestamp: () => findEndTimestamp,
   getChapterDuration: () => getChapterDuration,
   getSentenceRanges: () => getSentenceRanges,
-  interpolateSentenceRanges: () => interpolateSentenceRanges,
   mapTranscriptionTimeline: () => mapTranscriptionTimeline
 });
 module.exports = __toCommonJS(getSentenceRanges_exports);
@@ -311,62 +310,6 @@ async function getSentenceRanges(transcriptionText, mappedTimeline, sentences, c
     lastFoundSentence
   };
 }
-async function getLargestGap(trailing, leading) {
-  const leadingGap = leading.start;
-  const trailingGap = await (0, import_ffmpeg.getTrackDuration)(trailing.audiofile) - trailing.end;
-  if (trailingGap > leadingGap) return [trailingGap, trailing.audiofile];
-  return [leadingGap, leading.audiofile];
-}
-async function interpolateSentenceRanges(sentenceRanges, chapterSentenceCounts) {
-  const interpolated = [];
-  for (let i = 0; i < sentenceRanges.length; i++) {
-    const endRange = sentenceRanges[i];
-    const startRange = sentenceRanges[i - 1] ?? {
-      id: 0,
-      audiofile: endRange.audiofile,
-      chapterId: endRange.chapterId,
-      start: 0,
-      end: 0
-    };
-    const newChapter = startRange.chapterId !== endRange.chapterId;
-    const newAudiofile = startRange.audiofile !== endRange.audiofile;
-    const count = newChapter ? chapterSentenceCounts[startRange.chapterId] - startRange.id - 1 : endRange.id - startRange.id - 1;
-    if (count === 0) {
-      interpolated.push(endRange);
-      continue;
-    }
-    let [diff, audiofile] = newAudiofile ? await getLargestGap(startRange, endRange) : [endRange.start - startRange.end, endRange.audiofile];
-    if (diff <= 0) {
-      if (newAudiofile) {
-        const rangeLength = endRange.end - endRange.start;
-        diff = rangeLength < 0.5 ? rangeLength / 2 : 0.25;
-        endRange.start = diff;
-      } else {
-        diff = 0.25;
-        startRange.end = startRange.start - diff;
-      }
-    }
-    const interpolatedLength = diff / count;
-    const start = newAudiofile ? 0 : startRange.end;
-    for (let i2 = 0; i2 < count; i2++) {
-      let id = startRange.id + i2 + 1;
-      let chapterId = startRange.chapterId;
-      if (newChapter && i2 > chapterSentenceCounts[startRange.chapterId] - startRange.id) {
-        id = i2;
-        chapterId = endRange.chapterId;
-      }
-      interpolated.push({
-        id,
-        chapterId,
-        start: start + interpolatedLength * i2,
-        end: start + interpolatedLength * (i2 + 1),
-        audiofile
-      });
-    }
-    interpolated.push(endRange);
-  }
-  return interpolated;
-}
 function expandEmptySentenceRanges(sentenceRanges) {
   const expandedRanges = [];
   for (const sentenceRange of sentenceRanges) {
@@ -418,6 +361,5 @@ function getChapterDuration(sentenceRanges) {
   findEndTimestamp,
   getChapterDuration,
   getSentenceRanges,
-  interpolateSentenceRanges,
   mapTranscriptionTimeline
 });

package/dist/align/getSentenceRanges.d.cts

@@ -54,7 +54,6 @@ declare function getSentenceRanges(transcriptionText: string, mappedTimeline: Ma
     firstFoundSentence: number;
     lastFoundSentence: number;
 }>;
-declare function interpolateSentenceRanges(sentenceRanges: SentenceRange[], chapterSentenceCounts: Record<string, number>): Promise<SentenceRange[]>;
 /**
  * Whisper sometimes provides words with no time information,
  * or start and end timestamps that are equal. EpubCheck complains
@@ -65,4 +64,4 @@ declare function expandEmptySentenceRanges<Range extends SentenceRange | WordRan
 declare function collapseSentenceRangeGaps(sentenceRanges: SentenceRange[]): Promise<SentenceRange[]>;
 declare function getChapterDuration(sentenceRanges: SentenceRange[]): number;
 
-export { type MappedTimeline, type SentenceRange, type StorytellerTimelineEntry, type StorytellerTranscription, type WordRange, collapseSentenceRangeGaps, expandEmptySentenceRanges, findEndTimestamp, getChapterDuration, getSentenceRanges, interpolateSentenceRanges, mapTranscriptionTimeline };
+export { type MappedTimeline, type SentenceRange, type StorytellerTimelineEntry, type StorytellerTranscription, type WordRange, collapseSentenceRangeGaps, expandEmptySentenceRanges, findEndTimestamp, getChapterDuration, getSentenceRanges, mapTranscriptionTimeline };

package/dist/align/getSentenceRanges.d.ts

@@ -54,7 +54,6 @@ declare function getSentenceRanges(transcriptionText: string, mappedTimeline: Ma
     firstFoundSentence: number;
     lastFoundSentence: number;
 }>;
-declare function interpolateSentenceRanges(sentenceRanges: SentenceRange[], chapterSentenceCounts: Record<string, number>): Promise<SentenceRange[]>;
 /**
  * Whisper sometimes provides words with no time information,
  * or start and end timestamps that are equal. EpubCheck complains
@@ -65,4 +64,4 @@ declare function expandEmptySentenceRanges<Range extends SentenceRange | WordRan
 declare function collapseSentenceRangeGaps(sentenceRanges: SentenceRange[]): Promise<SentenceRange[]>;
 declare function getChapterDuration(sentenceRanges: SentenceRange[]): number;
 
-export { type MappedTimeline, type SentenceRange, type StorytellerTimelineEntry, type StorytellerTranscription, type WordRange, collapseSentenceRangeGaps, expandEmptySentenceRanges, findEndTimestamp, getChapterDuration, getSentenceRanges, interpolateSentenceRanges, mapTranscriptionTimeline };
+export { type MappedTimeline, type SentenceRange, type StorytellerTimelineEntry, type StorytellerTranscription, type WordRange, collapseSentenceRangeGaps, expandEmptySentenceRanges, findEndTimestamp, getChapterDuration, getSentenceRanges, mapTranscriptionTimeline };

package/dist/align/getSentenceRanges.js

@@ -283,62 +283,6 @@ async function getSentenceRanges(transcriptionText, mappedTimeline, sentences, c
     lastFoundSentence
   };
 }
-async function getLargestGap(trailing, leading) {
-  const leadingGap = leading.start;
-  const trailingGap = await getTrackDuration(trailing.audiofile) - trailing.end;
-  if (trailingGap > leadingGap) return [trailingGap, trailing.audiofile];
-  return [leadingGap, leading.audiofile];
-}
-async function interpolateSentenceRanges(sentenceRanges, chapterSentenceCounts) {
-  const interpolated = [];
-  for (let i = 0; i < sentenceRanges.length; i++) {
-    const endRange = sentenceRanges[i];
-    const startRange = sentenceRanges[i - 1] ?? {
-      id: 0,
-      audiofile: endRange.audiofile,
-      chapterId: endRange.chapterId,
-      start: 0,
-      end: 0
-    };
-    const newChapter = startRange.chapterId !== endRange.chapterId;
-    const newAudiofile = startRange.audiofile !== endRange.audiofile;
-    const count = newChapter ? chapterSentenceCounts[startRange.chapterId] - startRange.id - 1 : endRange.id - startRange.id - 1;
-    if (count === 0) {
-      interpolated.push(endRange);
-      continue;
-    }
-    let [diff, audiofile] = newAudiofile ? await getLargestGap(startRange, endRange) : [endRange.start - startRange.end, endRange.audiofile];
-    if (diff <= 0) {
-      if (newAudiofile) {
-        const rangeLength = endRange.end - endRange.start;
-        diff = rangeLength < 0.5 ? rangeLength / 2 : 0.25;
-        endRange.start = diff;
-      } else {
-        diff = 0.25;
-        startRange.end = startRange.start - diff;
-      }
-    }
-    const interpolatedLength = diff / count;
-    const start = newAudiofile ? 0 : startRange.end;
-    for (let i2 = 0; i2 < count; i2++) {
-      let id = startRange.id + i2 + 1;
-      let chapterId = startRange.chapterId;
-      if (newChapter && i2 > chapterSentenceCounts[startRange.chapterId] - startRange.id) {
-        id = i2;
-        chapterId = endRange.chapterId;
-      }
-      interpolated.push({
-        id,
-        chapterId,
-        start: start + interpolatedLength * i2,
-        end: start + interpolatedLength * (i2 + 1),
-        audiofile
-      });
-    }
-    interpolated.push(endRange);
-  }
-  return interpolated;
-}
 function expandEmptySentenceRanges(sentenceRanges) {
   const expandedRanges = [];
   for (const sentenceRange of sentenceRanges) {
@@ -389,6 +333,5 @@ export {
   findEndTimestamp,
   getChapterDuration,
   getSentenceRanges,
-  interpolateSentenceRanges,
   mapTranscriptionTimeline
 };

package/dist/align/interpolateSentenceRanges.cjs

@@ -0,0 +1,124 @@
+"use strict";
+var __defProp = Object.defineProperty;
+var __getOwnPropDesc = Object.getOwnPropertyDescriptor;
+var __getOwnPropNames = Object.getOwnPropertyNames;
+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 __toCommonJS = (mod) => __copyProps(__defProp({}, "__esModule", { value: true }), mod);
+var interpolateSentenceRanges_exports = {};
+__export(interpolateSentenceRanges_exports, {
+  interpolateSentenceRanges: () => interpolateSentenceRanges
+});
+module.exports = __toCommonJS(interpolateSentenceRanges_exports);
+function buildGapRanges(slots, left, right, audioFileDurations) {
+  const n = slots.length;
+  if (n === 0) return [];
+  if (left.audiofile === right.audiofile) {
+    const span = right.time - left.time;
+    return slots.map((slot, i) => ({
+      ...slot,
+      audiofile: left.audiofile,
+      start: left.time + span * i / n,
+      end: left.time + span * (i + 1) / n
+    }));
+  }
+  const leftDuration = audioFileDurations[left.audiofile] ?? left.time;
+  const leftAvail = leftDuration - left.time;
+  const rightAvail = right.time;
+  const total = leftAvail + rightAvail;
+  let n1 = total > 0 ? Math.round(n * (leftAvail / total)) : n;
+  let n2 = n - n1;
+  n1 = Math.max(0, n1);
+  n2 = n - n1;
+  const result = [];
+  if (n1 > 0) {
+    for (let i = 0; i < n1; i++) {
+      result.push({
+        // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+        ...slots[i],
+        audiofile: left.audiofile,
+        start: left.time + leftAvail * i / n1,
+        end: left.time + leftAvail * (i + 1) / n1
+      });
+    }
+  }
+  if (n2 > 0) {
+    for (let i = 0; i < n2; i++) {
+      result.push({
+        // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+        ...slots[n1 + i],
+        audiofile: right.audiofile,
+        start: rightAvail * i / n2,
+        end: rightAvail * (i + 1) / n2
+      });
+    }
+  }
+  return result;
+}
+function interpolateSentenceRanges(sentenceRanges, chapterSentenceCounts, audioFileDurations) {
+  if (sentenceRanges.length === 0) return [];
+  const result = [];
+  const first = sentenceRanges[0];
+  if (first.id > 0) {
+    const slots = Array.from({ length: first.id }, (_, i) => ({
+      chapterId: first.chapterId,
+      id: i
+    }));
+    const left = { time: 0, audiofile: first.audiofile };
+    const right = { time: first.start, audiofile: first.audiofile };
+    result.push(...buildGapRanges(slots, left, right, audioFileDurations));
+  }
+  result.push(first);
+  for (let idx = 1; idx < sentenceRanges.length; idx++) {
+    const prev = sentenceRanges[idx - 1];
+    const curr = sentenceRanges[idx];
+    const left = { time: prev.end, audiofile: prev.audiofile };
+    const right = { time: curr.start, audiofile: curr.audiofile };
+    const gapSlots = [];
+    if (prev.chapterId === curr.chapterId) {
+      for (let id = prev.id + 1; id < curr.id; id++) {
+        gapSlots.push({ chapterId: prev.chapterId, id });
+      }
+    } else {
+      const prevTotal = chapterSentenceCounts[prev.chapterId] ?? prev.id + 1;
+      for (let id = prev.id + 1; id < prevTotal; id++) {
+        gapSlots.push({ chapterId: prev.chapterId, id });
+      }
+      for (let id = 0; id < curr.id; id++) {
+        gapSlots.push({ chapterId: curr.chapterId, id });
+      }
+    }
+    if (gapSlots.length > 0) {
+      result.push(...buildGapRanges(gapSlots, left, right, audioFileDurations));
+    }
+    result.push(curr);
+  }
+  const last = sentenceRanges[sentenceRanges.length - 1];
+  const lastTotal = chapterSentenceCounts[last.chapterId] ?? last.id + 1;
+  if (last.id < lastTotal - 1) {
+    const slots = Array.from(
+      { length: lastTotal - 1 - last.id },
+      (_, i) => ({ chapterId: last.chapterId, id: last.id + 1 + i })
+    );
+    const fileEnd = audioFileDurations[last.audiofile] ?? last.end;
+    const left = { time: last.end, audiofile: last.audiofile };
+    const right = { time: fileEnd, audiofile: last.audiofile };
+    result.push(...buildGapRanges(slots, left, right, audioFileDurations));
+  }
+  return result;
+}
+// Annotate the CommonJS export names for ESM import in node:
+0 && (module.exports = {
+  interpolateSentenceRanges
+});

package/dist/align/interpolateSentenceRanges.d.cts

@@ -0,0 +1,23 @@
+import { SentenceRange } from './getSentenceRanges.cjs';
+import '@storyteller-platform/ghost-story';
+import '@echogarden/text-segmentation';
+import '@storyteller-platform/transliteration';
+
+/**
+ * Given a sequence of sentence ranges from an entire book,
+ * ordered by occurrence in audio, interpolates sentence ranges
+ * to fill any gaps.
+ *
+ * A gap may be:
+ *   - A non-linearity between two sequential sentence ranges
+ *     in the same chapter, e.g. chapter001#0 -> chapter001#3
+ *   - A chapter whose sentence ranges start at a number greater
+ *     than 0, e.g. chapter001#330 -> chapter002#2
+ *   - A chapter whose sentence ranges end at a number lower
+ *     than the total number of sentences in that chapter,
+ *     e.g. chapter001#325 -> chapter002#0, where
+ *     chapterSentenceCounts["chapter001"] === 330
+ */
+declare function interpolateSentenceRanges(sentenceRanges: SentenceRange[], chapterSentenceCounts: Record<string, number>, audioFileDurations: Record<string, number>): SentenceRange[];
+
+export { interpolateSentenceRanges };

package/dist/align/interpolateSentenceRanges.d.ts

@@ -0,0 +1,23 @@
+import { SentenceRange } from './getSentenceRanges.js';
+import '@storyteller-platform/ghost-story';
+import '@echogarden/text-segmentation';
+import '@storyteller-platform/transliteration';
+
+/**
+ * Given a sequence of sentence ranges from an entire book,
+ * ordered by occurrence in audio, interpolates sentence ranges
+ * to fill any gaps.
+ *
+ * A gap may be:
+ *   - A non-linearity between two sequential sentence ranges
+ *     in the same chapter, e.g. chapter001#0 -> chapter001#3
+ *   - A chapter whose sentence ranges start at a number greater
+ *     than 0, e.g. chapter001#330 -> chapter002#2
+ *   - A chapter whose sentence ranges end at a number lower
+ *     than the total number of sentences in that chapter,
+ *     e.g. chapter001#325 -> chapter002#0, where
+ *     chapterSentenceCounts["chapter001"] === 330
+ */
+declare function interpolateSentenceRanges(sentenceRanges: SentenceRange[], chapterSentenceCounts: Record<string, number>, audioFileDurations: Record<string, number>): SentenceRange[];
+
+export { interpolateSentenceRanges };

package/dist/align/interpolateSentenceRanges.js

@@ -0,0 +1,101 @@
+import "../chunk-BIEQXUOY.js";
+function buildGapRanges(slots, left, right, audioFileDurations) {
+  const n = slots.length;
+  if (n === 0) return [];
+  if (left.audiofile === right.audiofile) {
+    const span = right.time - left.time;
+    return slots.map((slot, i) => ({
+      ...slot,
+      audiofile: left.audiofile,
+      start: left.time + span * i / n,
+      end: left.time + span * (i + 1) / n
+    }));
+  }
+  const leftDuration = audioFileDurations[left.audiofile] ?? left.time;
+  const leftAvail = leftDuration - left.time;
+  const rightAvail = right.time;
+  const total = leftAvail + rightAvail;
+  let n1 = total > 0 ? Math.round(n * (leftAvail / total)) : n;
+  let n2 = n - n1;
+  n1 = Math.max(0, n1);
+  n2 = n - n1;
+  const result = [];
+  if (n1 > 0) {
+    for (let i = 0; i < n1; i++) {
+      result.push({
+        // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+        ...slots[i],
+        audiofile: left.audiofile,
+        start: left.time + leftAvail * i / n1,
+        end: left.time + leftAvail * (i + 1) / n1
+      });
+    }
+  }
+  if (n2 > 0) {
+    for (let i = 0; i < n2; i++) {
+      result.push({
+        // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
+        ...slots[n1 + i],
+        audiofile: right.audiofile,
+        start: rightAvail * i / n2,
+        end: rightAvail * (i + 1) / n2
+      });
+    }
+  }
+  return result;
+}
+function interpolateSentenceRanges(sentenceRanges, chapterSentenceCounts, audioFileDurations) {
+  if (sentenceRanges.length === 0) return [];
+  const result = [];
+  const first = sentenceRanges[0];
+  if (first.id > 0) {
+    const slots = Array.from({ length: first.id }, (_, i) => ({
+      chapterId: first.chapterId,
+      id: i
+    }));
+    const left = { time: 0, audiofile: first.audiofile };
+    const right = { time: first.start, audiofile: first.audiofile };
+    result.push(...buildGapRanges(slots, left, right, audioFileDurations));
+  }
+  result.push(first);
+  for (let idx = 1; idx < sentenceRanges.length; idx++) {
+    const prev = sentenceRanges[idx - 1];
+    const curr = sentenceRanges[idx];
+    const left = { time: prev.end, audiofile: prev.audiofile };
+    const right = { time: curr.start, audiofile: curr.audiofile };
+    const gapSlots = [];
+    if (prev.chapterId === curr.chapterId) {
+      for (let id = prev.id + 1; id < curr.id; id++) {
+        gapSlots.push({ chapterId: prev.chapterId, id });
+      }
+    } else {
+      const prevTotal = chapterSentenceCounts[prev.chapterId] ?? prev.id + 1;
+      for (let id = prev.id + 1; id < prevTotal; id++) {
+        gapSlots.push({ chapterId: prev.chapterId, id });
+      }
+      for (let id = 0; id < curr.id; id++) {
+        gapSlots.push({ chapterId: curr.chapterId, id });
+      }
+    }
+    if (gapSlots.length > 0) {
+      result.push(...buildGapRanges(gapSlots, left, right, audioFileDurations));
+    }
+    result.push(curr);
+  }
+  const last = sentenceRanges[sentenceRanges.length - 1];
+  const lastTotal = chapterSentenceCounts[last.chapterId] ?? last.id + 1;
+  if (last.id < lastTotal - 1) {
+    const slots = Array.from(
+      { length: lastTotal - 1 - last.id },
+      (_, i) => ({ chapterId: last.chapterId, id: last.id + 1 + i })
+    );
+    const fileEnd = audioFileDurations[last.audiofile] ?? last.end;
+    const left = { time: last.end, audiofile: last.audiofile };
+    const right = { time: fileEnd, audiofile: last.audiofile };
+    result.push(...buildGapRanges(slots, left, right, audioFileDurations));
+  }
+  return result;
+}
+export {
+  interpolateSentenceRanges
+};

package/dist/align/search.cjs

@@ -37,16 +37,16 @@ function buildNgramIndex(text) {
   }
   return index;
 }
+const NGRAM_SIZE = 5;
 function* ngrams(text) {
   const words = text.split("-");
-  let pos = 0;
-  for (const i of (0, import_itertools.range)(words.length - 4)) {
-    const ngram = words.slice(i, i + 5).join("-");
-    yield [ngram, pos];
-    pos += words[i].length + 1;
+  for (const i of (0, import_itertools.range)(words.length - NGRAM_SIZE - 1)) {
+    const ngram = words.slice(i, i + NGRAM_SIZE).join("-");
+    yield [ngram, i];
   }
 }
 function collectBoundaryVotes(query, document) {
+  const queryWords = query.split("-");
   const documentIndex = buildNgramIndex(document);
   let skippedNgrams = 0;
   let totalNgrams = 0;
@@ -61,7 +61,7 @@ function collectBoundaryVotes(query, document) {
     }
     for (const documentStart of documentStarts) {
       startVotes.push(documentStart - start);
-      endVotes.push(documentStart + (query.length - start));
+      endVotes.push(documentStart + (queryWords.length - start));
     }
   }
   if (skippedNgrams > totalNgrams / 2) {
@@ -97,6 +97,14 @@ function chooseBestFromBins(bins, dir) {
   }
   return dir > 0 ? (0, import_itertools.max)(best) ?? null : (0, import_itertools.min)(best) ?? null;
 }
+function getOffsetFromWordIndex(wordIndex, document) {
+  const words = document.split("-");
+  let offset = 0;
+  for (const i of (0, import_itertools.range)(Math.min(words.length, Math.max(0, wordIndex)))) {
+    offset += words[i].length + 1;
+  }
+  return offset;
+}
 function findBoundaries(query, document) {
   const boundaryVotes = collectBoundaryVotes(query, document);
   if (!boundaryVotes) return null;
@@ -111,7 +119,10 @@ function findBoundaries(query, document) {
   if (bestEnd === null) {
     return null;
   }
-  return { start: bestStart, end: bestEnd };
+  return {
+    start: getOffsetFromWordIndex(bestStart, document),
+    end: getOffsetFromWordIndex(bestEnd, document)
+  };
 }
 // Annotate the CommonJS export names for ESM import in node:
 0 && (module.exports = {