Remove the need for mp3 intermediate conversion

src/audio_splitter.py

@@ -14,36 +14,26 @@ def split(
     max_dur: float = 15,
     max_silence: float = 0.5,
     energy_threshold: float = 50,
-    expand_segments_with_noise: bool = False,
     noise_seconds: int = 1,
     noise_amplitude: int = 0,
-  ) -> list[tuple[str, float, float]]:
-    segments = split(
-      file_path,
-      min_dur=min_dur,
-      max_dur=max_dur,
-      max_silence=max_silence,
-      energy_threshold=energy_threshold,
-    )
-
-    if expand_segments_with_noise:
-      segments = [
-        (
-          self._expand_segment_with_noise(segment, noise_seconds, noise_amplitude),
-          segment.meta.start,
-          segment.meta.end,
-        ) for segment in segments
-      ]
-
-    return self._semgnets_to_data(segments)
-
-  def _expand_segment_with_noise(
-    self,
-    segment: AudioRegion,
-    noise_seconds: int,
-    noise_amplitude: int,
-  ) -> AudioSegment:
-
+  ) -> list[tuple[bytes, float, float]]:
+    segments = [
+      (
+        self._expand_segment_with_noise(segment, noise_seconds, noise_amplitude),
+        segment.meta.start,
+        segment.meta.end,
+      ) for segment in split(
+        file_path,
+        min_dur=min_dur,
+        max_dur=max_dur,
+        max_silence=max_silence,
+        energy_threshold=energy_threshold,
+      )
+    ]
+
+    return self._segments_to_data(segments)
+
+  def _expand_segment_with_noise(self, segment: AudioRegion, noise_seconds: int, noise_amplitude: int) -> AudioSegment:
     audio_segment = AudioSegment(
       segment._data,
       frame_rate=segment.sampling_rate,
@@ -56,20 +46,13 @@ def _expand_segment_with_noise(
 
     return pre_noise + audio_segment + post_noise
 
-  def _semgnets_to_data(
-    self,
-    segments: list[AudioSegment | tuple[AudioSegment, float, float]],
-  ) -> list[tuple[bytes, float, float]]:
-    segment_data = []
+  def _segments_to_data(self, segments: list[tuple[AudioSegment, float, float]]) -> list[tuple[bytes, float, float]]:
+    segments_data = []
 
     for segment in segments:
       output_buffer = io.BytesIO()
 
-      if isinstance(segment, tuple):
-        segment[0].export(output_buffer, format='mp3')
-        segment_data.append((output_buffer.getvalue(), segment[1], segment[2]))
-      else:
-        segment.export(output_buffer, format='mp3')
-        segment_data.append((output_buffer.getvalue(), segment.meta.start, segment.meta.end))
+      segment[0].export(output_buffer, format='mp3')
+      segments_data.append((output_buffer.getvalue(), segment[1], segment[2]))
 
-    return segment_data
+    return segments_data

src/cli.py

@@ -22,7 +22,6 @@
   import requests
 
   from .recognizers.wit_recognizer import WitRecognizer
-  from .utils.wit import file_utils as wit_file_utils
 except ModuleNotFoundError:
   pass
 
@@ -150,8 +149,7 @@ def process_local(
     file_path = str(file['file_path'].absolute())
 
     if config.use_wit():
-      mp3_file_path = str(wit_file_utils.convert_to_mp3(file['file_path']).absolute())
-      recognize_generator = WitRecognizer(verbose=config.input.verbose).recognize(mp3_file_path, config.wit)
+      recognize_generator = WitRecognizer(verbose=config.input.verbose).recognize(file_path, config.wit)
     else:
       recognize_generator = WhisperRecognizer(verbose=config.input.verbose).recognize(
           file_path,
@@ -167,9 +165,6 @@ def process_local(
         segments: list[SegmentType] = exception.value
         break
 
-    if config.use_wit() and file['file_path'].suffix != '.mp3':
-      Path(mp3_file_path).unlink(missing_ok=True)
-
     writer.write_all(Path(file['file_name']).stem, segments, config.output)
 
     for segment in segments:
@@ -212,13 +207,13 @@ def process_url(
     yield new_progress_info, []
 
     writer = Writer()
-    if config.input.skip_if_output_exist and writer.is_output_exist(element['id'], config.output):
+    if config.input.skip_if_output_exist and writer.is_output_exist(element['id'], element['audio_ext']):
       new_progress_info['inner_status'] = 'completed'
       yield new_progress_info, []
 
       continue
 
-    file_path = os.path.join(config.output.output_dir, f"{element['id']}.mp3")
+    file_path = os.path.join(config.output.output_dir, f"{element['id']}.{element['audio_ext']}")
 
     if config.use_wit():
       recognize_generator = WitRecognizer(verbose=config.input.verbose).recognize(file_path, config.wit)

src/downloader.py

@@ -23,12 +23,6 @@ def _config(self, download_archive: str | bool) -> dict[str, Any]:
       'ignoreerrors': True,
       'download_archive': download_archive,
       'playlist_items': self.playlist_items,
-      'postprocessors': [
-        {
-          'key': 'FFmpegExtractAudio',
-          'preferredcodec': 'mp3',
-        },
-      ],
     }
 
   def download(self, url: str, save_response: bool = False) -> dict[str, Any]:

src/recognizers/wit_recognizer.py

@@ -33,11 +33,7 @@ def recognize(
     file_path: str,
     wit_config: Config.Wit,
   ) -> Generator[dict[str, float], None, list[SegmentType]]:
-    segments = AudioSplitter().split(
-      file_path,
-      max_dur=wit_config.max_cutting_duration,
-      expand_segments_with_noise=True,
-    )
+    segments = AudioSplitter().split(file_path, max_dur=wit_config.max_cutting_duration)
 
     retry_strategy = Retry(
       total=5,