fix

之前的错误会导致合并字幕 word 出现问题
Huanshere · Sep 12, 2024 · c6617f4 · c6617f4
1 parent f019913
commit c6617f4
Show file tree

Hide file tree

Showing 2 changed files with 14 additions and 7 deletions.
diff --git a/core/all_whisper_methods/whisperX.py b/core/all_whisper_methods/whisperX.py
@@ -70,28 +70,32 @@ def transcribe_audio(audio_file: str) -> Dict:
         raise Exception(f"WhisperX processing error: {e}")
 
 def process_transcription(result: Dict) -> pd.DataFrame:
+    from config import get_joiner, WHISPER_LANGUAGE
+    language = result['language'] if WHISPER_LANGUAGE == 'auto' else WHISPER_LANGUAGE # consider force english case
+    joiner = get_joiner(language)
+
     all_words = []
     for segment in result['segments']:
         for word in segment['words']:
+            # ! For French, we need to convert guillemets to empty strings
+            word["word"] = word["word"].replace('»', '').replace('«', '')
+
             if 'start' not in word and 'end' not in word:
                 if all_words:
                     # Merge with the previous word
-                    all_words[-1]['text'] = f'{all_words[-1]["text"][:-1]}{word["word"]}"'
+                    all_words[-1]['text'] = f'{all_words[-1]["text"]}{joiner}{word["word"]}'
                 else:
                     # If it's the first word, temporarily save it and wait for the next word with a timestamp
                     temp_word = word["word"]
             else:
                 # Normal case, with start and end times
                 word_dict = {
-                    'text': f'"{temp_word}{word["word"]}"' if 'temp_word' in locals() else f'"{word["word"]}"',
+                    'text': f'{temp_word}{word["word"]}' if 'temp_word' in locals() else f'{word["word"]}',
                     'start': word.get('start', all_words[-1]['end'] if all_words else 0),
                     'end': word['end'],
                     'score': word.get('score', 0)
                 }
 
-                # ! For French, we need to convert guillemets to empty strings
-                word_dict['text'] = word_dict['text'].replace('»', '').replace('«', '')
-
                 all_words.append(word_dict)
                 if 'temp_word' in locals():
                     del temp_word

diff --git a/core/all_whisper_methods/whisperXapi.py b/core/all_whisper_methods/whisperXapi.py
@@ -66,8 +66,11 @@ def transcribe_audio(audio_base64: str) -> Dict:
     except Exception as e:
         raise Exception(f"Error accessing whisperX API: {e} Please check your Replicate API key and internet connection.\n")
 
-
 def process_transcription(result: Dict) -> pd.DataFrame:
+    from config import get_joiner, WHISPER_LANGUAGE
+    language = result['detected_language'] if WHISPER_LANGUAGE == 'auto' else WHISPER_LANGUAGE # consider force english case
+    joiner = get_joiner(language)
+
     all_words = []
     for segment in result['segments']:
         for word in segment['words']:
@@ -77,7 +80,7 @@ def process_transcription(result: Dict) -> pd.DataFrame:
             if 'start' not in word and 'end' not in word:
                 if all_words:
                     # Merge with the previous word
-                    all_words[-1]['text'] = f'{all_words[-1]["text"][:-1]}{word["word"]}"'
+                    all_words[-1]['text'] = f'{all_words[-1]["text"]}{joiner}{word["word"]}'
                 else:
                     # If it's the first word, temporarily save it and wait for the next word with a timestamp
                     temp_word = word["word"]