@storyteller-platform/align 0.1.36 → 0.1.38

package/dist/align/align.cjs

@@ -105,14 +105,25 @@ async function align(input, output, transcriptionsDir, audiobookDir, options) {
   var _stack2 = [];
   try {
     const outFormat = options.outFormat ?? "epub";
+    const epubPath = outFormat === "epub" ? (0, import_node_path.join)(
+      (0, import_node_os.tmpdir)(),
+      `storyteller-platform-align-${(0, import_node_crypto.randomUUID)()}`,
+      (0, import_posix.basename)(output)
+    ) : input;
+    const stack = __using(_stack2, new DisposableStack());
+    stack.defer(() => {
+      if (outFormat === "epub") {
+        (0, import_node_fs.rmSync)((0, import_posix.dirname)(epubPath), { recursive: true, force: true });
+      }
+    });
     if (outFormat === "epub") {
-      await (0, import_promises.mkdir)((0, import_posix.dirname)(output), { recursive: true });
-      await (0, import_promises.copyFile)(input, output);
+      await (0, import_promises.mkdir)((0, import_posix.dirname)(epubPath), { recursive: true });
+      await (0, import_promises.copyFile)(input, epubPath);
     }
     const audiobookFiles = await (0, import_promises.readdir)(audiobookDir).then(
       (filenames) => filenames.filter((f) => (0, import_audiobook.isAudioFile)(f)).map((f) => (0, import_node_path.join)(audiobookDir, f))
     );
-    const epub = __using(_stack2, await import_epub.Epub.from(outFormat === "epub" ? output : input));
+    const epub = __using(_stack2, await import_epub.Epub.from(epubPath));
     const transcriptions = await (0, import_promises.readdir)(transcriptionsDir).then(
       (filenames) => filenames.filter((f) => f.endsWith(".json")).map((f) => (0, import_node_path.join)(transcriptionsDir, f))
     ).then(
@@ -146,6 +157,8 @@ async function align(input, output, transcriptionsDir, audiobookDir, options) {
     const timing = await aligner.alignBook(options.onProgress);
     if (outFormat === "epub") {
       await epub.saveAndClose();
+      await (0, import_promises.mkdir)((0, import_posix.dirname)(output), { recursive: true });
+      await (0, import_promises.copyFile)(epubPath, output);
     } else {
       var _stack = [];
       try {
@@ -166,8 +179,8 @@ async function align(input, output, transcriptionsDir, audiobookDir, options) {
         writeStream.on("close", () => {
           resolve();
         });
-        const stack = __using(_stack, new AsyncDisposableStack(), true);
-        stack.defer(async () => {
+        const stack2 = __using(_stack, new AsyncDisposableStack(), true);
+        stack2.defer(async () => {
           writeStream.close();
           await (0, import_promises.rm)(tmpArchivePath, { force: true });
         });
@@ -229,8 +242,12 @@ class Aligner {
   timing = (0, import_ghost_story.createAggregator)();
   granularity;
   textRef;
+  audioFileDurations = {};
   report = {
-    chapters: []
+    chapters: [],
+    unalignedChapters: [],
+    audioFiles: [],
+    unalignedAudioFiles: []
   };
   async getChapterSentences(chapterId) {
     const chapterXml = await this.epub.readXhtmlItemContents(chapterId);
@@ -301,18 +318,24 @@ class Aligner {
           );
         }
         if (this.granularity === "word") {
-          const wordFactory = new import_textFragments.TextFragmentFactory(
-            blockRanges.flatMap((range) => {
-              const sentence = sentences[range.id];
-              const wordRanges2 = wordRangeMap.get(range.id);
-              const toFragment = wordIdToFragment.get(range.id);
-              if (!wordRanges2 || !toFragment) return [];
-              const words = sentence.words.entries.filter(
-                (w) => w.text.match(/\S/)
-              );
-              return words.map((w) => w.text.replace("\n", ""));
-            })
-          );
+          const allWords = [];
+          for (const range of blockRanges) {
+            const sentence = sentences[range.id];
+            const words = [];
+            for (const w of sentence.words.entries) {
+              if (w.isPunctuation) {
+                const lastWord = words.at(-1);
+                if (lastWord === void 0) {
+                  continue;
+                }
+                words[words.length - 1] = lastWord + w.text.replace("\n", "");
+              } else {
+                words.push(w.text);
+              }
+            }
+            allWords.push(...words);
+          }
+          const wordFactory = new import_textFragments.TextFragmentFactory(allWords);
           let wordRangeIndex = 0;
           for (const range of blockRanges) {
             const wordRanges2 = wordRangeMap.get(range.id);
@@ -388,19 +411,53 @@ class Aligner {
       value: import_epub.Epub.formatSmilDuration(chapterDuration)
     });
   }
-  addChapterReport(chapter, chapterSentences, sentenceRanges, startSentence, endSentence, transcriptionOffset) {
+  addChapterReport(chapter, chapterSentences, sentenceRanges, startSentence, endSentence, mapping, transcriptionOffset, endTranscriptionOffset) {
+    const audioFiles = sentenceRanges.reduce(
+      (acc, range) => {
+        const existing = acc.find(
+          (context) => context.filepath === range.audiofile
+        );
+        if (existing) {
+          existing.end = range.end;
+          return acc;
+        }
+        acc.push({
+          filepath: range.audiofile,
+          start: range.start,
+          end: range.end
+        });
+        return acc;
+      },
+      []
+    );
+    const mappedTranscriptionOffset = mapping.invert().map(transcriptionOffset);
+    const mappedEndTranscriptionOffset = mapping.invert().map(endTranscriptionOffset);
     this.report.chapters.push({
       href: chapter.href,
-      transcriptionOffset,
+      transcriptionOffset: mappedTranscriptionOffset,
+      endTranscriptionOffset: mappedEndTranscriptionOffset,
       transcriptionContext: {
         before: this.transcription.transcript.slice(
-          Math.max(0, transcriptionOffset - 30),
-          transcriptionOffset
+          Math.max(0, mappedTranscriptionOffset - 80),
+          mappedTranscriptionOffset
         ),
         after: this.transcription.transcript.slice(
-          transcriptionOffset,
+          mappedTranscriptionOffset,
           Math.min(
-            transcriptionOffset + 30,
+            mappedTranscriptionOffset + 80,
+            this.transcription.transcript.length - 1
+          )
+        )
+      },
+      endTranscriptionContext: {
+        before: this.transcription.transcript.slice(
+          Math.max(0, mappedEndTranscriptionOffset - 80),
+          mappedEndTranscriptionOffset
+        ),
+        after: this.transcription.transcript.slice(
+          mappedEndTranscriptionOffset,
+          Math.min(
+            mappedEndTranscriptionOffset + 80,
             this.transcription.transcript.length - 1
           )
         )
@@ -421,24 +478,30 @@ class Aligner {
       },
       chapterSentenceCount: chapterSentences.length,
       alignedSentenceCount: sentenceRanges.length,
-      audioFiles: sentenceRanges.reduce((acc, range) => {
-        const existing = acc.find(
-          (context) => context.filepath === range.audiofile
-        );
-        if (existing) {
-          existing.end = range.end;
-          return acc;
-        }
-        acc.push({
-          filepath: range.audiofile,
-          start: range.start,
-          end: range.end
-        });
-        return acc;
-      }, [])
+      audioFiles
     });
+    for (const audioFile of audioFiles) {
+      const existing = this.report.audioFiles.find(
+        ({ filepath }) => audioFile.filepath === filepath
+      );
+      if (existing) {
+        existing.matchedRanges.push({
+          start: audioFile.start,
+          end: audioFile.end
+        });
+        existing.matchedRanges.sort((a, b) => a.start - b.start);
+        existing.alignedDuration += audioFile.end - audioFile.start;
+      } else {
+        this.report.audioFiles.push({
+          alignedDuration: audioFile.end - audioFile.start,
+          duration: this.audioFileDurations[audioFile.filepath] ?? 0,
+          filepath: audioFile.filepath,
+          matchedRanges: [{ start: audioFile.start, end: audioFile.end }]
+        });
+      }
+    }
   }
-  async alignChapter(chapterId, transcriptionText, transcriptionOffset, transcriptionEndOffset, locale, mappedTimeline) {
+  async alignChapter(chapterId, transcriptionText, transcriptionOffset, transcriptionEndOffset, locale, mappedTimeline, mapping) {
     const timing = (0, import_ghost_story.createTiming)();
     timing.start("read contents");
     const manifest = await this.epub.getManifest();
@@ -493,7 +556,9 @@ class Aligner {
       sentenceRanges,
       firstFoundSentence,
       lastFoundSentence,
-      transcriptionOffset
+      mapping,
+      transcriptionOffset,
+      endTranscriptionOffset
     );
     return {
       lastSentenceRange: sentenceRanges.at(-1) ?? null,
@@ -525,6 +590,9 @@ class Aligner {
     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);
+    for (const audiofile of this.audiofiles) {
+      this.audioFileDurations[audiofile] = await (0, import_ffmpeg.getTrackDuration)(audiofile);
+    }
     const spine = await this.epub.getSpineItems();
     const manifest = await this.epub.getManifest();
     const { result: transcriptionText, mapping } = await (0, import_slugify.slugify)(
@@ -540,6 +608,10 @@ class Aligner {
       );
       const chapterId = spineItem.id;
       if (manifest[chapterId]?.properties?.includes("nav")) {
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "is-nav"
+        });
         continue;
       }
       const chapterSentences = await this.getChapterSentences(chapterId);
@@ -551,6 +623,10 @@ class Aligner {
       }
       if (chapterSentences.length === 0) {
         this.logger?.info(`Chapter #${index} has no text; skipping`);
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "no-text"
+        });
         continue;
       }
       if (chapterSentences.length < 2 && // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
@@ -558,6 +634,10 @@ class Aligner {
         this.logger?.info(
           `Chapter #${index} is fewer than four words; skipping`
         );
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "too-short"
+        });
         continue;
       }
       const boundaries = (0, import_search.findBoundaries)(
@@ -568,6 +648,12 @@ class Aligner {
         this.logger?.info(
           `Could not find chapter #${index} in the transcripton`
         );
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "not-found",
+          start: chapterSentences.slice(0, 3).map((s) => s.text).join("").slice(0, 80),
+          end: chapterSentences.slice(-3).map((s) => s.text).join("").slice(-80)
+        });
         continue;
       }
       const { start, end } = this.narrowToAvailableBoundary(boundaries);
@@ -580,7 +666,8 @@ class Aligner {
         Math.max(start, 0),
         Math.min(end, transcriptionText.length),
         locale,
-        mappedTimeline
+        mappedTimeline,
+        mapping
       );
       this.timing.add(result.timing.summary());
     }
@@ -602,16 +689,8 @@ 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
       );
@@ -620,7 +699,7 @@ class Aligner {
     const interpolated = (0, import_interpolateSentenceRanges.interpolateSentenceRanges)(
       sentenceRanges,
       chapterSentenceCounts,
-      audioFileDurations
+      this.audioFileDurations
     );
     const expanded = (0, import_getSentenceRanges.expandEmptySentenceRanges)(interpolated);
     const collapsed = await (0, import_getSentenceRanges.collapseSentenceRangeGaps)(expanded);
@@ -640,6 +719,11 @@ class Aligner {
       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" },

package/dist/align/align.d.cts

@@ -14,10 +14,15 @@ interface AudioFileContext {
 interface ChapterReport {
     href: string;
     transcriptionOffset: number;
+    endTranscriptionOffset: number;
     transcriptionContext: {
         before: string;
         after: string;
     };
+    endTranscriptionContext: {
+        before: string;
+        after: string;
+    };
     firstMatchedSentenceId: number;
     firstMatchedSentenceContext: {
         prevSentence: string | null;
@@ -34,8 +39,34 @@ interface ChapterReport {
     alignedSentenceCount: number;
     audioFiles: AudioFileContext[];
 }
+type UnalignedChapterReason = "too-short" | "not-found" | "is-nav" | "no-text";
+interface UnalignedChapterReport {
+    href: string;
+    reason: Exclude<UnalignedChapterReason, "not-found">;
+}
+interface UnalignedNotFoundChapterReport {
+    href: string;
+    reason: "not-found";
+    start: string;
+    end: string;
+}
+interface AudioFileReport {
+    filepath: string;
+    matchedRanges: {
+        start: number;
+        end: number;
+    }[];
+    duration: number;
+    alignedDuration: number;
+}
+interface UnalignedAudioFileReport {
+    filepath: string;
+}
 interface Report {
     chapters: ChapterReport[];
+    unalignedChapters: (UnalignedChapterReport | UnalignedNotFoundChapterReport)[];
+    audioFiles: AudioFileReport[];
+    unalignedAudioFiles: UnalignedAudioFileReport[];
 }
 interface AlignOptions {
     reportsPath?: string | null | undefined;
@@ -58,6 +89,7 @@ declare class Aligner {
     private timing;
     private granularity;
     private textRef;
+    private audioFileDurations;
     report: Report;
     constructor(epub: Epub, audiofiles: string[], transcriptions: Pick<RecognitionResult, "transcript" | "timeline">[], granularity: "sentence" | "word" | null | undefined, textRef: "id-fragment" | "text-fragment" | null | undefined, languageOverride?: (Intl.Locale | null) | undefined, logger?: (Logger | null) | undefined);
     private getChapterSentences;

package/dist/align/align.d.ts

@@ -14,10 +14,15 @@ interface AudioFileContext {
 interface ChapterReport {
     href: string;
     transcriptionOffset: number;
+    endTranscriptionOffset: number;
     transcriptionContext: {
         before: string;
         after: string;
     };
+    endTranscriptionContext: {
+        before: string;
+        after: string;
+    };
     firstMatchedSentenceId: number;
     firstMatchedSentenceContext: {
         prevSentence: string | null;
@@ -34,8 +39,34 @@ interface ChapterReport {
     alignedSentenceCount: number;
     audioFiles: AudioFileContext[];
 }
+type UnalignedChapterReason = "too-short" | "not-found" | "is-nav" | "no-text";
+interface UnalignedChapterReport {
+    href: string;
+    reason: Exclude<UnalignedChapterReason, "not-found">;
+}
+interface UnalignedNotFoundChapterReport {
+    href: string;
+    reason: "not-found";
+    start: string;
+    end: string;
+}
+interface AudioFileReport {
+    filepath: string;
+    matchedRanges: {
+        start: number;
+        end: number;
+    }[];
+    duration: number;
+    alignedDuration: number;
+}
+interface UnalignedAudioFileReport {
+    filepath: string;
+}
 interface Report {
     chapters: ChapterReport[];
+    unalignedChapters: (UnalignedChapterReport | UnalignedNotFoundChapterReport)[];
+    audioFiles: AudioFileReport[];
+    unalignedAudioFiles: UnalignedAudioFileReport[];
 }
 interface AlignOptions {
     reportsPath?: string | null | undefined;
@@ -58,6 +89,7 @@ declare class Aligner {
     private timing;
     private granularity;
     private textRef;
+    private audioFileDurations;
     report: Report;
     constructor(epub: Epub, audiofiles: string[], transcriptions: Pick<RecognitionResult, "transcript" | "timeline">[], granularity: "sentence" | "word" | null | undefined, textRef: "id-fragment" | "text-fragment" | null | undefined, languageOverride?: (Intl.Locale | null) | undefined, logger?: (Logger | null) | undefined);
     private getChapterSentences;

package/dist/align/align.js

@@ -3,7 +3,7 @@ import {
   __using
 } from "../chunk-BIEQXUOY.js";
 import { randomUUID } from "node:crypto";
-import { createWriteStream } from "node:fs";
+import { createWriteStream, rmSync } from "node:fs";
 import {
   copyFile,
   cp,
@@ -51,14 +51,25 @@ async function align(input, output, transcriptionsDir, audiobookDir, options) {
   var _stack2 = [];
   try {
     const outFormat = options.outFormat ?? "epub";
+    const epubPath = outFormat === "epub" ? autoJoin(
+      tmpdir(),
+      `storyteller-platform-align-${randomUUID()}`,
+      basename(output)
+    ) : input;
+    const stack = __using(_stack2, new DisposableStack());
+    stack.defer(() => {
+      if (outFormat === "epub") {
+        rmSync(dirname(epubPath), { recursive: true, force: true });
+      }
+    });
     if (outFormat === "epub") {
-      await mkdir(dirname(output), { recursive: true });
-      await copyFile(input, output);
+      await mkdir(dirname(epubPath), { recursive: true });
+      await copyFile(input, epubPath);
     }
     const audiobookFiles = await readdir(audiobookDir).then(
       (filenames) => filenames.filter((f) => isAudioFile(f)).map((f) => autoJoin(audiobookDir, f))
     );
-    const epub = __using(_stack2, await Epub.from(outFormat === "epub" ? output : input));
+    const epub = __using(_stack2, await Epub.from(epubPath));
     const transcriptions = await readdir(transcriptionsDir).then(
       (filenames) => filenames.filter((f) => f.endsWith(".json")).map((f) => autoJoin(transcriptionsDir, f))
     ).then(
@@ -92,6 +103,8 @@ async function align(input, output, transcriptionsDir, audiobookDir, options) {
     const timing = await aligner.alignBook(options.onProgress);
     if (outFormat === "epub") {
       await epub.saveAndClose();
+      await mkdir(dirname(output), { recursive: true });
+      await copyFile(epubPath, output);
     } else {
       var _stack = [];
       try {
@@ -112,8 +125,8 @@ async function align(input, output, transcriptionsDir, audiobookDir, options) {
         writeStream.on("close", () => {
           resolve();
         });
-        const stack = __using(_stack, new AsyncDisposableStack(), true);
-        stack.defer(async () => {
+        const stack2 = __using(_stack, new AsyncDisposableStack(), true);
+        stack2.defer(async () => {
           writeStream.close();
           await rm(tmpArchivePath, { force: true });
         });
@@ -175,8 +188,12 @@ class Aligner {
   timing = createAggregator();
   granularity;
   textRef;
+  audioFileDurations = {};
   report = {
-    chapters: []
+    chapters: [],
+    unalignedChapters: [],
+    audioFiles: [],
+    unalignedAudioFiles: []
   };
   async getChapterSentences(chapterId) {
     const chapterXml = await this.epub.readXhtmlItemContents(chapterId);
@@ -247,18 +264,24 @@ class Aligner {
           );
         }
         if (this.granularity === "word") {
-          const wordFactory = new TextFragmentFactory(
-            blockRanges.flatMap((range) => {
-              const sentence = sentences[range.id];
-              const wordRanges2 = wordRangeMap.get(range.id);
-              const toFragment = wordIdToFragment.get(range.id);
-              if (!wordRanges2 || !toFragment) return [];
-              const words = sentence.words.entries.filter(
-                (w) => w.text.match(/\S/)
-              );
-              return words.map((w) => w.text.replace("\n", ""));
-            })
-          );
+          const allWords = [];
+          for (const range of blockRanges) {
+            const sentence = sentences[range.id];
+            const words = [];
+            for (const w of sentence.words.entries) {
+              if (w.isPunctuation) {
+                const lastWord = words.at(-1);
+                if (lastWord === void 0) {
+                  continue;
+                }
+                words[words.length - 1] = lastWord + w.text.replace("\n", "");
+              } else {
+                words.push(w.text);
+              }
+            }
+            allWords.push(...words);
+          }
+          const wordFactory = new TextFragmentFactory(allWords);
           let wordRangeIndex = 0;
           for (const range of blockRanges) {
             const wordRanges2 = wordRangeMap.get(range.id);
@@ -334,19 +357,53 @@ class Aligner {
       value: Epub.formatSmilDuration(chapterDuration)
     });
   }
-  addChapterReport(chapter, chapterSentences, sentenceRanges, startSentence, endSentence, transcriptionOffset) {
+  addChapterReport(chapter, chapterSentences, sentenceRanges, startSentence, endSentence, mapping, transcriptionOffset, endTranscriptionOffset) {
+    const audioFiles = sentenceRanges.reduce(
+      (acc, range) => {
+        const existing = acc.find(
+          (context) => context.filepath === range.audiofile
+        );
+        if (existing) {
+          existing.end = range.end;
+          return acc;
+        }
+        acc.push({
+          filepath: range.audiofile,
+          start: range.start,
+          end: range.end
+        });
+        return acc;
+      },
+      []
+    );
+    const mappedTranscriptionOffset = mapping.invert().map(transcriptionOffset);
+    const mappedEndTranscriptionOffset = mapping.invert().map(endTranscriptionOffset);
     this.report.chapters.push({
       href: chapter.href,
-      transcriptionOffset,
+      transcriptionOffset: mappedTranscriptionOffset,
+      endTranscriptionOffset: mappedEndTranscriptionOffset,
       transcriptionContext: {
         before: this.transcription.transcript.slice(
-          Math.max(0, transcriptionOffset - 30),
-          transcriptionOffset
+          Math.max(0, mappedTranscriptionOffset - 80),
+          mappedTranscriptionOffset
         ),
         after: this.transcription.transcript.slice(
-          transcriptionOffset,
+          mappedTranscriptionOffset,
           Math.min(
-            transcriptionOffset + 30,
+            mappedTranscriptionOffset + 80,
+            this.transcription.transcript.length - 1
+          )
+        )
+      },
+      endTranscriptionContext: {
+        before: this.transcription.transcript.slice(
+          Math.max(0, mappedEndTranscriptionOffset - 80),
+          mappedEndTranscriptionOffset
+        ),
+        after: this.transcription.transcript.slice(
+          mappedEndTranscriptionOffset,
+          Math.min(
+            mappedEndTranscriptionOffset + 80,
             this.transcription.transcript.length - 1
           )
         )
@@ -367,24 +424,30 @@ class Aligner {
       },
       chapterSentenceCount: chapterSentences.length,
       alignedSentenceCount: sentenceRanges.length,
-      audioFiles: sentenceRanges.reduce((acc, range) => {
-        const existing = acc.find(
-          (context) => context.filepath === range.audiofile
-        );
-        if (existing) {
-          existing.end = range.end;
-          return acc;
-        }
-        acc.push({
-          filepath: range.audiofile,
-          start: range.start,
-          end: range.end
-        });
-        return acc;
-      }, [])
+      audioFiles
     });
+    for (const audioFile of audioFiles) {
+      const existing = this.report.audioFiles.find(
+        ({ filepath }) => audioFile.filepath === filepath
+      );
+      if (existing) {
+        existing.matchedRanges.push({
+          start: audioFile.start,
+          end: audioFile.end
+        });
+        existing.matchedRanges.sort((a, b) => a.start - b.start);
+        existing.alignedDuration += audioFile.end - audioFile.start;
+      } else {
+        this.report.audioFiles.push({
+          alignedDuration: audioFile.end - audioFile.start,
+          duration: this.audioFileDurations[audioFile.filepath] ?? 0,
+          filepath: audioFile.filepath,
+          matchedRanges: [{ start: audioFile.start, end: audioFile.end }]
+        });
+      }
+    }
   }
-  async alignChapter(chapterId, transcriptionText, transcriptionOffset, transcriptionEndOffset, locale, mappedTimeline) {
+  async alignChapter(chapterId, transcriptionText, transcriptionOffset, transcriptionEndOffset, locale, mappedTimeline, mapping) {
     const timing = createTiming();
     timing.start("read contents");
     const manifest = await this.epub.getManifest();
@@ -439,7 +502,9 @@ class Aligner {
       sentenceRanges,
       firstFoundSentence,
       lastFoundSentence,
-      transcriptionOffset
+      mapping,
+      transcriptionOffset,
+      endTranscriptionOffset
     );
     return {
       lastSentenceRange: sentenceRanges.at(-1) ?? null,
@@ -471,6 +536,9 @@ class Aligner {
     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);
+    for (const audiofile of this.audiofiles) {
+      this.audioFileDurations[audiofile] = await getTrackDuration(audiofile);
+    }
     const spine = await this.epub.getSpineItems();
     const manifest = await this.epub.getManifest();
     const { result: transcriptionText, mapping } = await slugify(
@@ -486,6 +554,10 @@ class Aligner {
       );
       const chapterId = spineItem.id;
       if (manifest[chapterId]?.properties?.includes("nav")) {
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "is-nav"
+        });
         continue;
       }
       const chapterSentences = await this.getChapterSentences(chapterId);
@@ -497,6 +569,10 @@ class Aligner {
       }
       if (chapterSentences.length === 0) {
         this.logger?.info(`Chapter #${index} has no text; skipping`);
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "no-text"
+        });
         continue;
       }
       if (chapterSentences.length < 2 && // eslint-disable-next-line @typescript-eslint/no-non-null-assertion
@@ -504,6 +580,10 @@ class Aligner {
         this.logger?.info(
           `Chapter #${index} is fewer than four words; skipping`
         );
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "too-short"
+        });
         continue;
       }
       const boundaries = findBoundaries(
@@ -514,6 +594,12 @@ class Aligner {
         this.logger?.info(
           `Could not find chapter #${index} in the transcripton`
         );
+        this.report.unalignedChapters.push({
+          href: spineItem.href,
+          reason: "not-found",
+          start: chapterSentences.slice(0, 3).map((s) => s.text).join("").slice(0, 80),
+          end: chapterSentences.slice(-3).map((s) => s.text).join("").slice(-80)
+        });
         continue;
       }
       const { start, end } = this.narrowToAvailableBoundary(boundaries);
@@ -526,7 +612,8 @@ class Aligner {
         Math.max(start, 0),
         Math.min(end, transcriptionText.length),
         locale,
-        mappedTimeline
+        mappedTimeline,
+        mapping
       );
       this.timing.add(result.timing.summary());
     }
@@ -548,16 +635,8 @@ 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
       );
@@ -566,7 +645,7 @@ class Aligner {
     const interpolated = interpolateSentenceRanges(
       sentenceRanges,
       chapterSentenceCounts,
-      audioFileDurations
+      this.audioFileDurations
     );
     const expanded = expandEmptySentenceRanges(interpolated);
     const collapsed = await collapseSentenceRangeGaps(expanded);
@@ -586,6 +665,11 @@ class Aligner {
       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" },

package/dist/align/parse.cjs

@@ -40,7 +40,9 @@ const alignParser = (0, import_core.object)("Alignment", {
     }),
     "id-fragment"
   ),
-  reports: (0, import_core.optional)((0, import_core.option)("--reports", (0, import_valueparser.path)({ type: "directory" }))),
+  reports: (0, import_core.optional)(
+    (0, import_core.option)("--reports", (0, import_valueparser.path)({ type: "file", extensions: [".json"] }))
+  ),
   outFormat: (0, import_core.withDefault)(
     (0, import_core.option)("--out-format", (0, import_core.choice)(["epub", "gnp"]), {
       description: import_core.message`Whether to output a full EPUB 3 package with embedded media overlays and audio, or a Readium Guided Navigation Package with just a manifest and guided navigation documents.`

package/dist/align/parse.js

@@ -31,7 +31,9 @@ const alignParser = object("Alignment", {
     }),
     "id-fragment"
   ),
-  reports: optional(option("--reports", path({ type: "directory" }))),
+  reports: optional(
+    option("--reports", path({ type: "file", extensions: [".json"] }))
+  ),
   outFormat: withDefault(
     option("--out-format", choice(["epub", "gnp"]), {
       description: message`Whether to output a full EPUB 3 package with embedded media overlays and audio, or a Readium Guided Navigation Package with just a manifest and guided navigation documents.`

package/dist/align/slugify.cjs

@@ -133,6 +133,7 @@ async function slugify(text, locale) {
   replacerMap.set(locale, replacers);
   const { result, mapping } = await (0, import_transliteration.slugify)(text, {
     allowedChars: "a-zA-Z0-9",
+    locale,
     replace: replacers
   });
   return { result, mapping };

package/dist/align/slugify.js

@@ -111,6 +111,7 @@ async function slugify(text, locale) {
   replacerMap.set(locale, replacers);
   const { result, mapping } = await transliterateSlugify(text, {
     allowedChars: "a-zA-Z0-9",
+    locale,
     replace: replacers
   });
   return { result, mapping };

package/dist/align/textFragments.cjs

@@ -64,15 +64,23 @@ class TextFragmentFactory {
       toRemove.toReversed().map((r) => candidates.splice(r, 1));
       i++;
     }
+    while (chars.at(i)?.match(/[\p{L}\p{N}]/u) && i < chars.length) i++;
     let fragment = "";
     const start = chars.slice(0, i).join("");
     fragment += encodeTextFragmentPart(start);
-    const remainingSpan = span.slice(i);
+    const remainingChars = chars.slice(i);
+    while (!remainingChars.at(-1)?.match(/[\p{L}\p{N}]/u) && remainingChars.length) {
+      remainingChars.splice(remainingChars.length - 1, 1);
+    }
+    let e = remainingChars.length;
     let end = "";
-    let e = remainingSpan.length - 1;
-    if (remainingSpan.at(-1) === "\n") e--;
-    while (remainingSpan.indexOf(end) !== e + 1 && e >= 0) {
-      end = remainingSpan.slice(e);
+    const remainingSpan = remainingChars.join("");
+    while (remainingSpan.indexOf(end) !== remainingSpan.length - remainingChars.slice(e).join("").length && e >= 0) {
+      e--;
+      end = remainingChars.slice(e).join("");
+    }
+    while (remainingChars.at(e)?.match(/[\p{L}\p{N}]/u) && e >= 0) {
+      end = remainingChars.slice(e).join("");
       e--;
     }
     if (end) {
@@ -93,7 +101,10 @@ class TextFragmentFactory {
         p++;
         if (!candidates.length) break;
       }
-      const prefix = this.runes.slice(startPos - p + 1, startPos).join("");
+      while (this.runes.at(startPos - p - 1)?.match(/[\p{L}\p{N}]/u) && p <= startPos) {
+        p++;
+      }
+      const prefix = this.runes.slice(startPos - p, startPos).join("");
       fragment = `${encodeTextFragmentPart(prefix)}-,${fragment}`;
     }
     return `:~:text=${fragment}`;

package/dist/align/textFragments.js

@@ -42,15 +42,23 @@ class TextFragmentFactory {
       toRemove.toReversed().map((r) => candidates.splice(r, 1));
       i++;
     }
+    while (chars.at(i)?.match(/[\p{L}\p{N}]/u) && i < chars.length) i++;
     let fragment = "";
     const start = chars.slice(0, i).join("");
     fragment += encodeTextFragmentPart(start);
-    const remainingSpan = span.slice(i);
+    const remainingChars = chars.slice(i);
+    while (!remainingChars.at(-1)?.match(/[\p{L}\p{N}]/u) && remainingChars.length) {
+      remainingChars.splice(remainingChars.length - 1, 1);
+    }
+    let e = remainingChars.length;
     let end = "";
-    let e = remainingSpan.length - 1;
-    if (remainingSpan.at(-1) === "\n") e--;
-    while (remainingSpan.indexOf(end) !== e + 1 && e >= 0) {
-      end = remainingSpan.slice(e);
+    const remainingSpan = remainingChars.join("");
+    while (remainingSpan.indexOf(end) !== remainingSpan.length - remainingChars.slice(e).join("").length && e >= 0) {
+      e--;
+      end = remainingChars.slice(e).join("");
+    }
+    while (remainingChars.at(e)?.match(/[\p{L}\p{N}]/u) && e >= 0) {
+      end = remainingChars.slice(e).join("");
       e--;
     }
     if (end) {
@@ -71,7 +79,10 @@ class TextFragmentFactory {
         p++;
         if (!candidates.length) break;
       }
-      const prefix = this.runes.slice(startPos - p + 1, startPos).join("");
+      while (this.runes.at(startPos - p - 1)?.match(/[\p{L}\p{N}]/u) && p <= startPos) {
+        p++;
+      }
+      const prefix = this.runes.slice(startPos - p, startPos).join("");
       fragment = `${encodeTextFragmentPart(prefix)}-,${fragment}`;
     }
     return `:~:text=${fragment}`;

package/dist/cli/bin.cjs

@@ -229,6 +229,7 @@ async function main() {
             textRef: parsed.textRef,
             outFormat: parsed.outFormat,
             primaryLocale: parsed.language,
+            reportsPath: parsed.reports,
             logger,
             ...!parsed.noProgress && parsed.logLevel === "silent" && {
               onProgress: (progress) => {
@@ -349,6 +350,7 @@ async function main() {
               textRef: parsed.textRef,
               outFormat: parsed.outFormat,
               primaryLocale,
+              reportsPath: parsed.reports,
               logger,
               ...!parsed.noProgress && parsed.logLevel === "silent" && {
                 onProgress: (progress) => {

package/dist/cli/bin.js

@@ -180,6 +180,7 @@ async function main() {
             textRef: parsed.textRef,
             outFormat: parsed.outFormat,
             primaryLocale: parsed.language,
+            reportsPath: parsed.reports,
             logger,
             ...!parsed.noProgress && parsed.logLevel === "silent" && {
               onProgress: (progress) => {
@@ -300,6 +301,7 @@ async function main() {
               textRef: parsed.textRef,
               outFormat: parsed.outFormat,
               primaryLocale,
+              reportsPath: parsed.reports,
               logger,
               ...!parsed.noProgress && parsed.logLevel === "silent" && {
                 onProgress: (progress) => {

package/dist/markup/markup.cjs

@@ -68,6 +68,7 @@ __export(markup_exports, {
 });
 module.exports = __toCommonJS(markup_exports);
 var import_promises = require("node:fs/promises");
+var import_node_path = require("node:path");
 var import_posix = require("node:path/posix");
 var import_epub = require("@storyteller-platform/epub");
 var import_ghost_story = require("@storyteller-platform/ghost-story");
@@ -81,6 +82,7 @@ async function markup(input, output, options) {
   try {
     const timing = (0, import_ghost_story.createAggregator)();
     timing.setMetadata("granularity", options.granularity ?? "sentence");
+    await (0, import_promises.mkdir)((0, import_node_path.dirname)(output), { recursive: true });
     await (0, import_promises.copyFile)(input, output);
     const epub = __using(_stack, await import_epub.Epub.from(output));
     const primaryLocale = options.primaryLocale ?? await epub.getLanguage();

package/dist/markup/markup.js

@@ -2,7 +2,8 @@ import {
   __callDispose,
   __using
 } from "../chunk-BIEQXUOY.js";
-import { copyFile } from "node:fs/promises";
+import { copyFile, mkdir } from "node:fs/promises";
+import { dirname } from "node:path";
 import { basename } from "node:path/posix";
 import { Epub } from "@storyteller-platform/epub";
 import {
@@ -24,6 +25,7 @@ async function markup(input, output, options) {
   try {
     const timing = createAggregator();
     timing.setMetadata("granularity", options.granularity ?? "sentence");
+    await mkdir(dirname(output), { recursive: true });
     await copyFile(input, output);
     const epub = __using(_stack, await Epub.from(output));
     const primaryLocale = options.primaryLocale ?? await epub.getLanguage();

package/dist/markup/model.d.cts

@@ -1,2 +1,2 @@
 import '@storyteller-platform/epub';
-export { F as FootnoteNode, M as Mark, N as Node, b as NoterefNode, R as Root, T as TextNode, d as descendants } from '../model-TZi1QUQh.cjs';
+export { F as FootnoteNode, M as Mark, N as Node, b as NoterefNode, R as Root, T as TextNode, d as descendants } from '../model-Bv3yPEdd.cjs';

package/dist/markup/model.d.ts

@@ -1,2 +1,2 @@
 import '@storyteller-platform/epub';
-export { F as FootnoteNode, M as Mark, N as Node, b as NoterefNode, R as Root, T as TextNode, d as descendants } from '../model-TZi1QUQh.js';
+export { F as FootnoteNode, M as Mark, N as Node, b as NoterefNode, R as Root, T as TextNode, d as descendants } from '../model-Bv3yPEdd.js';

package/dist/markup/parseDom.d.cts

@@ -1,5 +1,5 @@
 import { ParsedXml } from '@storyteller-platform/epub';
-import { R as Root, N as Node } from '../model-TZi1QUQh.cjs';
+import { R as Root, N as Node } from '../model-Bv3yPEdd.cjs';
 
 declare function parseDom(xml: ParsedXml): Root;
 declare function findFootnotePairs(root: Root | Node): Map<number, number>;

package/dist/markup/parseDom.d.ts

@@ -1,5 +1,5 @@
 import { ParsedXml } from '@storyteller-platform/epub';
-import { R as Root, N as Node } from '../model-TZi1QUQh.js';
+import { R as Root, N as Node } from '../model-Bv3yPEdd.js';
 
 declare function parseDom(xml: ParsedXml): Root;
 declare function findFootnotePairs(root: Root | Node): Map<number, number>;

package/dist/markup/resolvedPos.d.cts

@@ -1,2 +1,2 @@
-export { a as ResolvedPos } from '../model-TZi1QUQh.cjs';
+export { a as ResolvedPos } from '../model-Bv3yPEdd.cjs';
 import '@storyteller-platform/epub';

package/dist/markup/resolvedPos.d.ts

@@ -1,2 +1,2 @@
-export { a as ResolvedPos } from '../model-TZi1QUQh.js';
+export { a as ResolvedPos } from '../model-Bv3yPEdd.js';
 import '@storyteller-platform/epub';

package/dist/markup/serializeDom.d.cts

@@ -1,5 +1,5 @@
 import { ParsedXml } from '@storyteller-platform/epub';
-import { R as Root } from '../model-TZi1QUQh.cjs';
+import { R as Root } from '../model-Bv3yPEdd.cjs';
 
 declare function serializeDom(doc: Root): ParsedXml;
 

package/dist/markup/serializeDom.d.ts

@@ -1,5 +1,5 @@
 import { ParsedXml } from '@storyteller-platform/epub';
-import { R as Root } from '../model-TZi1QUQh.js';
+import { R as Root } from '../model-Bv3yPEdd.js';
 
 declare function serializeDom(doc: Root): ParsedXml;
 

package/dist/markup/transform.cjs

@@ -81,11 +81,22 @@ function liftText(root) {
         ])
       );
     }
-    lastTextEnd = pos + node.nodeSize;
     let result = node.text.replaceAll(/\n/g, " ");
+    if (text.endsWith("\n")) {
+      const contentStart = result.match(/\S/u)?.index ?? result.length;
+      if (contentStart !== 0) {
+        result = result.slice(contentStart);
+        mapping.appendMap(
+          new import_map.StepMap([mapping.map(lastTextEnd), contentStart, 0])
+        );
+      }
+    }
+    lastTextEnd = pos + node.nodeSize;
     const hasBlockSiblings = parent.children.some((child) => child.isBlock);
     if (hasBlockSiblings && !result.match(/\S/)) {
-      mapping.appendMap(new import_map.StepMap([textLength, result.length, 0]));
+      if (result.length) {
+        mapping.appendMap(new import_map.StepMap([textLength, result.length, 0]));
+      }
       result = "";
     }
     if (parent.isBlock && index === parent.children.length - 1 && !(text + result).endsWith("\n")) {

package/dist/markup/transform.d.cts

@@ -1,5 +1,5 @@
 import { Mapping } from './map.cjs';
-import { R as Root, M as Mark } from '../model-TZi1QUQh.cjs';
+import { R as Root, M as Mark } from '../model-Bv3yPEdd.cjs';
 import '@storyteller-platform/epub';
 
 declare function addMark(root: Root, from: number, to: number, mark: Mark): Root;

package/dist/markup/transform.d.ts

@@ -1,5 +1,5 @@
 import { Mapping } from './map.js';
-import { R as Root, M as Mark } from '../model-TZi1QUQh.js';
+import { R as Root, M as Mark } from '../model-Bv3yPEdd.js';
 import '@storyteller-platform/epub';
 
 declare function addMark(root: Root, from: number, to: number, mark: Mark): Root;

package/dist/markup/transform.js

@@ -61,11 +61,22 @@ function liftText(root) {
         ])
       );
     }
-    lastTextEnd = pos + node.nodeSize;
     let result = node.text.replaceAll(/\n/g, " ");
+    if (text.endsWith("\n")) {
+      const contentStart = result.match(/\S/u)?.index ?? result.length;
+      if (contentStart !== 0) {
+        result = result.slice(contentStart);
+        mapping.appendMap(
+          new StepMap([mapping.map(lastTextEnd), contentStart, 0])
+        );
+      }
+    }
+    lastTextEnd = pos + node.nodeSize;
     const hasBlockSiblings = parent.children.some((child) => child.isBlock);
     if (hasBlockSiblings && !result.match(/\S/)) {
-      mapping.appendMap(new StepMap([textLength, result.length, 0]));
+      if (result.length) {
+        mapping.appendMap(new StepMap([textLength, result.length, 0]));
+      }
       result = "";
     }
     if (parent.isBlock && index === parent.children.length - 1 && !(text + result).endsWith("\n")) {

package/dist/{model-TZi1QUQh.d.cts → model-Bv3yPEdd.d.cts}

@@ -45,7 +45,7 @@ declare class Node {
     get isLeaf(): boolean;
     get isInline(): boolean;
     get isBlock(): boolean;
-    get border(): 1 | 0;
+    get border(): 0 | 1;
     get nodeSize(): number;
     get contentSize(): number;
     get textContent(): string;

package/dist/{model-TZi1QUQh.d.ts → model-Bv3yPEdd.d.ts}

@@ -45,7 +45,7 @@ declare class Node {
     get isLeaf(): boolean;
     get isInline(): boolean;
     get isBlock(): boolean;
-    get border(): 1 | 0;
+    get border(): 0 | 1;
     get nodeSize(): number;
     get contentSize(): number;
     get textContent(): string;

package/dist/snapshot/snapshot.cjs

@@ -71,6 +71,7 @@ var import_promises = require("node:fs/promises");
 var import_node_path = require("node:path");
 var import_posix = require("node:path/posix");
 var import_epub = require("@storyteller-platform/epub");
+var import_ghost_story = require("@storyteller-platform/ghost-story");
 var import_parseDom = require("../markup/parseDom.cjs");
 var import_segmentation = require("../markup/segmentation.cjs");
 var import_transform = require("../markup/transform.cjs");
@@ -184,7 +185,9 @@ async function createAlignmentSnapshot(epub, transcriptionFilepaths, textRef) {
         }
         word = transcription.timeline[++i];
       }
-      const transcriptionSentence = transcriptionWords.join(" ");
+      const transcriptionSentence = transcriptionWords.map(
+        (w, idx) => (0, import_ghost_story.startsWithSpacelessScript)(w) || idx === transcriptionWords.length - 1 ? w : `${w} `
+      ).join("");
       newSnapshot += `Audio: ${transcriptionSentence}
 `;
     }

package/dist/snapshot/snapshot.js

@@ -11,6 +11,9 @@ import {
 import {
   Epub
 } from "@storyteller-platform/epub";
+import {
+  startsWithSpacelessScript
+} from "@storyteller-platform/ghost-story";
 import { parseDom } from "../markup/parseDom.js";
 import { segmentChapter } from "../markup/segmentation.js";
 import { inlineFootnotes, liftText } from "../markup/transform.js";
@@ -124,7 +127,9 @@ async function createAlignmentSnapshot(epub, transcriptionFilepaths, textRef) {
         }
         word = transcription.timeline[++i];
       }
-      const transcriptionSentence = transcriptionWords.join(" ");
+      const transcriptionSentence = transcriptionWords.map(
+        (w, idx) => startsWithSpacelessScript(w) || idx === transcriptionWords.length - 1 ? w : `${w} `
+      ).join("");
       newSnapshot += `Audio: ${transcriptionSentence}
 `;
     }

package/dist/transcribe/transcribe.cjs

@@ -84,10 +84,6 @@ var import_async_semaphore = require("@esfx/async-semaphore");
 var import_audiobook = require("@storyteller-platform/audiobook");
 var import_ghost_story = require("@storyteller-platform/ghost-story");
 async function transcribe(input, output, locale, options) {
-  if (process.env["DEBUG_TRANSCRIBE"] === "true") {
-    const inspector = await import("node:inspector");
-    inspector.open(9231, "0.0.0.0", true);
-  }
   const semaphore = new import_async_semaphore.AsyncSemaphore(options.parallelism ?? 1);
   const controller = new AbortController();
   const signal = AbortSignal.any([

package/dist/transcribe/transcribe.js

@@ -15,10 +15,6 @@ import {
   recognize
 } from "@storyteller-platform/ghost-story";
 async function transcribe(input, output, locale, options) {
-  if (process.env["DEBUG_TRANSCRIBE"] === "true") {
-    const inspector = await import("node:inspector");
-    inspector.open(9231, "0.0.0.0", true);
-  }
   const semaphore = new AsyncSemaphore(options.parallelism ?? 1);
   const controller = new AbortController();
   const signal = AbortSignal.any([

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@storyteller-platform/align",
-  "version": "0.1.36",
+  "version": "0.1.38",
   "description": "A library and CLI for automatically aligning audiobooks and EPUBs to produce Media Overlays",
   "author": "Shane Friedman",
   "license": "MIT",
@@ -62,7 +62,7 @@
     "@readium/shared": "patch:@readium/shared@npm%3A2.1.5#~/.yarn/patches/@readium-shared-npm-2.1.5-8d6f9d2432.patch",
     "@storyteller-platform/audiobook": "^0.3.10",
     "@storyteller-platform/epub": "^0.5.0",
-    "@storyteller-platform/ghost-story": "^0.1.10",
+    "@storyteller-platform/ghost-story": "^0.1.11",
     "@storyteller-platform/transliteration": "^3.1.2",
     "chalk": "^5.4.1",
     "change-case": "^5.4.4",