Merge pull request #418 from JenySadadia/enable-timestamps

Enable timestamps for transcripted text

run_client.py

@@ -47,6 +47,9 @@ if __name__ == '__main__':
                           type=str,
                           default='fr',
                           help='Target language for translation, e.g., "fr" for French.')
+    parser.add_argument('--enable_timestamps',
+                          action='store_true',
+                          help='Show transcription with timestamps')
 
     args = parser.parse_args()
 
@@ -80,5 +83,6 @@ if __name__ == '__main__':
             mute_audio_playback=args.mute_audio_playback,      # Only used for file input, False by Default
             enable_translation=args.enable_translation,        # Enable translation of the transcription output
             target_language=args.target_language,              # Target language for translation, e.g., "fr
+            enable_timestamps=args.enable_timestamps,
         )
         client(f)

whisper_live/client.py

@@ -41,6 +41,7 @@ class Client:
         target_language="fr",
         translation_callback=None,
         translation_srt_file_path="output_translated.srt",
+        enable_timestamps=False,
     ):
         """
         Initializes a Client instance for audio recording and streaming to a server.
@@ -97,6 +98,7 @@ class Client:
         self.last_translated_segment = None
         if translate:
             self.task = "translate"
+        self.enable_timestamps = enable_timestamps
 
         self.audio_bytes = None
 
@@ -175,17 +177,41 @@ class Client:
                 except Exception as e:
                     print(f"[WARN] transcription_callback raised: {e}")
                 return
-        
+
         if self.log_transcription:
-            original_text = [seg["text"] for seg in self.transcript[-4:]]
-            if self.last_segment is not None and self.last_segment["text"] not in original_text:
-                original_text.append(self.last_segment["text"])
-            
-            utils.clear_screen()
-            utils.print_transcript(original_text)
-            if self.enable_translation:
-                print(f"\n\nTRANSLATION to {self.target_language}:")
-                utils.print_transcript([seg["text"] for seg in self.translated_transcript[-4:]], translated=True)
+            if self.enable_timestamps:
+                original_text_with_timestamps = [
+                    {"start": seg["start"], "end": seg["end"], "text": seg["text"]}
+                    for seg in self.transcript[-4:]]
+                if self.last_segment is not None and not any(
+                    data.get("text") == self.last_segment["text"]
+                    for data in original_text_with_timestamps):
+                    original_text_with_timestamps.append({
+                        "start": self.last_segment["start"],
+                        "end": self.last_segment["end"],
+                        "text": self.last_segment["text"]
+                    })
+                utils.clear_screen()
+                utils.print_transcript(original_text_with_timestamps, timestamps=True)
+
+                if self.enable_translation:
+                    print(f"\n\nTRANSLATION to {self.target_language}:")
+                    utils.print_transcript([
+                        {"start": seg["start"], "end": seg["end"], "text": seg["text"]}
+                        for seg in self.translated_transcript[-4:]
+                    ], timestamps=True)
+
+            else:
+                original_text = [seg["text"] for seg in self.transcript[-4:]]
+                if self.last_segment is not None and self.last_segment["text"] not in original_text:
+                    original_text.append(self.last_segment["text"])
+                utils.clear_screen()
+                utils.print_transcript(original_text)
+
+                if self.enable_translation:
+                    print(f"\n\nTRANSLATION to {self.target_language}:")
+                    utils.print_transcript([seg["text"] for seg in self.translated_transcript[-4:]], translated=True)
+
 
     def on_message(self, ws, message):
         """
@@ -790,6 +816,7 @@ class TranscriptionClient(TranscriptionTeeClient):
         target_language="fr",
         translation_callback=None,
         translation_srt_file_path="./output_translated.srt",
+        enable_timestamps=False,
     ):
         self.client = Client(
             host,
@@ -810,6 +837,7 @@ class TranscriptionClient(TranscriptionTeeClient):
             target_language=target_language,
             translation_callback=translation_callback,
             translation_srt_file_path=translation_srt_file_path,
+            enable_timestamps=enable_timestamps,
         )
 
         if save_output_recording and not output_recording_filename.endswith(".wav"):

whisper_live/utils.py

@@ -11,12 +11,16 @@ def clear_screen():
     os.system("cls" if os.name == "nt" else "clear")
 
 
-def print_transcript(text, translated=False):
+def print_transcript(text, translated=False, timestamps=False):
     """Prints formatted transcript text."""
-    wrapper = textwrap.TextWrapper(width=60)
-    text=" ".join(text) if translated else "".join(text)
-    for line in wrapper.wrap(text=text):
-        print(line)
+    if timestamps:
+        for t in text:
+            print(f'[{t["start"]} -> {t["end"]}] {t["text"]}')
+    else:
+        wrapper = textwrap.TextWrapper(width=60)
+        text=" ".join(text) if translated else "".join(text)
+        for line in wrapper.wrap(text=text):
+            print(line)
 
 
 def format_time(s):