feat: progress callbacks

whisperx/alignment.py

@@ -119,6 +119,7 @@ def align(
     return_char_alignments: bool = False,
     print_progress: bool = False,
     combined_progress: bool = False,
+    progress_callback=None,
 ) -> AlignedTranscriptionResult:
     """
     Align phoneme recognition predictions to known transcription.
@@ -147,6 +148,8 @@ def align(
             base_progress = ((sdx + 1) / total_segments) * 100
             percent_complete = (50 + base_progress / 2) if combined_progress else base_progress
             print(f"Progress: {percent_complete:.2f}%...")
+            if progress_callback:
+                progress_callback(percent_complete)
 
         num_leading = len(segment["text"]) - len(segment["text"].lstrip())
         num_trailing = len(segment["text"]) - len(segment["text"].rstrip())

whisperx/asr.py

@@ -195,6 +195,7 @@ def transcribe(
         print_progress=False,
         combined_progress=False,
         verbose=False,
+        progress_callback=None,
     ) -> TranscriptionResult:
         if isinstance(audio, str):
             audio = load_audio(audio)
@@ -258,6 +259,8 @@ def data(audio, segments):
                 base_progress = ((idx + 1) / total_segments) * 100
                 percent_complete = base_progress / 2 if combined_progress else base_progress
                 print(f"Progress: {percent_complete:.2f}%...")
+                if progress_callback:
+                    progress_callback(percent_complete)
             text = out['text']
             if batch_size in [0, 1, None]:
                 text = text[0]
@@ -364,8 +367,8 @@ def load_model(
         "without_timestamps": True,
         "max_initial_timestamp": 0.0,
         "word_timestamps": False,
-        "prepend_punctuations": "\"'“¿([{-",
-        "append_punctuations": "\"'.。,，!！?？:：”)]}、",
+        "prepend_punctuations": "\"'¿([{-",
+        "append_punctuations": "\"'.。,，!！?？:："")]}、",
         "multilingual": model.model.is_multilingual,
         "suppress_numerals": False,
         "max_new_tokens": None,