feat: Add original text preservation in chunking and export processes

abogen/chunking.py

@@ -148,6 +148,7 @@ def chunk_text(
                 voice_formula=voice_formula,
             ).as_dict()
             payload["normalized_text"] = _normalize_chunk_text(paragraph)
+            payload["original_text"] = paragraph
             chunks.append(payload)
         _attach_display_text(text, chunks)
         return chunks
@@ -177,6 +178,7 @@ def chunk_text(
             ).as_dict()
             payload["normalized_text"] = _normalize_chunk_text(raw_sentence)
             payload["display_text"] = raw_sentence
+            payload["original_text"] = raw_sentence
             chunks.append(payload)
             sentence_index += 1
 
@@ -221,9 +223,11 @@ def _attach_display_text(source: str, chunks: List[Dict[str, object]]) -> None:
             match = _search_source_span(source, candidate, 0)
         if match is None:
             chunk.setdefault("display_text", candidate)
+            chunk.setdefault("original_text", chunk.get("display_text") or candidate)
             continue
         start, end = match
         chunk["display_text"] = source[start:end]
+        chunk["original_text"] = source[start:end]
         cursor = end
 
 

abogen/epub3/exporter.py

@@ -243,7 +243,7 @@ class EPUB3PackageBuilder:
                 voice = marker.get("voice")
             else:
                 display_text = chunk_entry.get("display_text")
-                text = str(display_text or chunk_entry.get("text") or "")
+                text = str(chunk_entry.get("text") or "")
                 speaker_id = str(chunk_entry.get("speaker_id") or marker.get("speaker_id") or "narrator")
                 voice = chunk_entry.get("voice") or chunk_entry.get("resolved_voice") or marker.get("voice")
                 level = chunk_entry.get("level") or None
@@ -261,11 +261,15 @@ class EPUB3PackageBuilder:
             group_id = _derive_group_id(raw_group_key, level)
             normalized_group_id = _normalize_chunk_id(group_id) if group_id else None
 
+            original_text = None
+            if chunk_entry is not None:
+                original_text = chunk_entry.get("original_text") or chunk_entry.get("display_text")
+
             overlays.append(
                 ChunkOverlay(
                     id=normalized_id,
                     text=text or self.extraction.chapters[chapter_index].text,
-                    original_text=None,
+                    original_text=str(original_text) if original_text is not None else None,
                     start=_safe_float(marker.get("start")),
                     end=_safe_float(marker.get("end")),
                     speaker_id=speaker_id,
@@ -747,6 +751,10 @@ def _restore_original_chunk_text(chapter_text: str, overlays: List[ChunkOverlay]
 
     cursor = 0
     for chunk in overlays:
+        if chunk.original_text is not None:
+            prepared = _prepare_display_text(chunk.original_text)
+            chunk.text = prepared
+            continue
         candidate = chunk.text or ""
         if not candidate:
             continue

abogen/web/service.py

@@ -1122,6 +1122,10 @@ class ConversionService:
             if normalized_value is not None:
                 chunk["normalized_text"] = str(normalized_value)
 
+            for text_key in ("display_text", "original_text"):
+                if text_key in entry and entry[text_key] is not None:
+                    chunk[text_key] = str(entry[text_key])
+
             speaker_value = entry.get("speaker_id", entry.get("speaker"))
             chunk["speaker_id"] = str(speaker_value) if speaker_value else "narrator"
 

tests/test_chunk_helpers.py

@@ -59,6 +59,8 @@ def test_chunk_text_merges_title_abbreviations() -> None:
     assert "Doctor" in normalized_value
     display_value = str(chunk.get("display_text") or "")
     assert display_value.startswith("Dr.")
+    original_value = str(chunk.get("original_text") or "")
+    assert original_value.startswith("Dr.")
 
 
 def test_chunk_text_display_preserves_whitespace() -> None:
@@ -77,3 +79,5 @@ def test_chunk_text_display_preserves_whitespace() -> None:
     assert first_display.endswith("\n\n")
     second_display = str(chunks[1].get("display_text") or "")
     assert second_display == "Third paragraph."
+    first_original = str(chunks[0].get("original_text") or "")
+    assert first_original.endswith("\n\n")