fix: timing/punctuation edge cases, process chunks as uncompressed wavs

1 files changed, 13 insertions, 5 deletions

diff --git a/converter/converter.py b/converter/converter.py
index f63a412..0454892 100644
--- a/converter/converter.py
+++ b/converter/converter.py
@@ -130,12 +130,17 @@ class AudiobookConverter:
 
     def _convert_m4b_with_chapters(self, sections, stem: str, start_time: float) -> bool:
         """Convert each chapter to audio, then assemble a single m4b with
-        embedded chapter markers."""
+        embedded chapter markers.
+
+        Chapters are synthesized to lossless WAV scratch files (~170 MB per
+        hour of audio) so the final AAC pass is the only lossy encode.
+        """
         chapter_files = []
         titles = []
         for index, section in enumerate(sections, 1):
-            chapter_path = config.CHUNKS_FOLDER / f"chapter_{index:04d}.{self.output_format}"
-            if not self._convert_text(section.text, chapter_path, start_time, speed=1.0):
+            chapter_path = config.CHUNKS_FOLDER / f"chapter_{index:04d}.wav"
+            if not self._convert_text(section.text, chapter_path, start_time,
+                                      speed=1.0, output_format="wav"):
                 logger.warning("Skipping chapter %d (%s) due to conversion failure",
                                index, section.title)
                 continue
@@ -150,10 +155,13 @@ class AudiobookConverter:
         return audio.combine_chapters_to_m4b(chapter_files, titles, output_path, speed=self.speed)
 
     def _convert_text(self, text: str, output_path: Path, start_time: float,
-                      speed: Optional[float] = None) -> bool:
+                      speed: Optional[float] = None,
+                      output_format: Optional[str] = None) -> bool:
         """Chunk, synthesize, and assemble ``text`` into ``output_path``."""
         if speed is None:
             speed = self.speed
+        if output_format is None:
+            output_format = self.output_format
 
         try:
             if not text.strip():
@@ -217,7 +225,7 @@ class AudiobookConverter:
 
             # Combine chunks (only the successful ones)
             success = audio.combine_chunks(total_chunks, output_path, results,
-                                           speed=speed, output_format=self.output_format)
+                                           speed=speed, output_format=output_format)
 
             if success:
                 duration = time.time() - start_time