Whisper Progress Bar - Display More Information

README.md

@@ -4,7 +4,7 @@ emoji: ✨
 colorFrom: blue
 colorTo: purple
 sdk: gradio
-sdk_version: 3.36.0
+sdk_version: 3.39.0
 app_file: app.py
 pinned: false
 license: apache-2.0

app.py

@@ -387,8 +387,8 @@ class WhisperTranscriber:
                 # From 0 to 1
                 self.progress(current / total, desc=desc)
 
-            def on_finished(self):
-                self.progress(1)
+            def on_finished(self, desc: str = None):
+                self.progress(1, desc=desc)
 
         return ForwardingProgressListener(progress)
 
@@ -466,13 +466,13 @@ class WhisperTranscriber:
                     if nllb_model.nllb_lang is not None:
                         segment["text"] = nllb_model.translation(seg_text)
                     pbar.update(1)
-                    segments_progress_listener.on_progress(idx+1, len(segments), "Process segments")
+                    segments_progress_listener.on_progress(idx+1, len(segments), desc=f"Process segments: {idx}/{len(segments)}")
 
                 nllb_model.release_vram()
                 perf_end_time = time.perf_counter()
                 # Call the finished callback
                 if segments_progress_listener is not None:
-                    segments_progress_listener.on_finished()
+                    segments_progress_listener.on_finished(desc=f"Process segments: {idx}/{len(segments)}")
 
                 print("\n\nprocess segments took {} seconds.\n\n".format(perf_end_time - perf_start_time))
             except Exception as e:

requirements-fasterWhisper.txt

@@ -2,7 +2,7 @@
 ctranslate2>=3.16.0
 faster-whisper
 ffmpeg-python==0.2.0
-gradio==3.36.0 
+gradio==3.39.0
 yt-dlp
 json5
 torch

requirements-whisper.txt

@@ -2,7 +2,7 @@
 git+https://github.com/openai/whisper.git
 transformers
 ffmpeg-python==0.2.0
-gradio==3.36.0
+gradio==3.39.0
 yt-dlp
 torchaudio
 altair

requirements.txt

@@ -2,7 +2,7 @@
 ctranslate2>=3.16.0
 faster-whisper
 ffmpeg-python==0.2.0
-gradio==3.36.0 
+gradio==3.39.0
 yt-dlp
 json5
 torch

src/hooks/progressListener.py

@@ -4,5 +4,5 @@ class ProgressListener:
     def on_progress(self, current: Union[int, float], total: Union[int, float], desc: str = None):
         self.total = total
 
-    def on_finished(self):
+    def on_finished(self, desc: str = None):
         pass

src/hooks/subTaskProgressListener.py

@@ -33,5 +33,5 @@ class SubTaskProgressListener(ProgressListener):
         sub_task_progress = self.sub_task_start + self.sub_task_total * sub_task_progress_frac
         self.base_task_listener.on_progress(sub_task_progress, self.base_task_total, desc=desc)
 
-    def on_finished(self):
-        self.base_task_listener.on_progress(self.sub_task_start + self.sub_task_total, self.base_task_total)
+    def on_finished(self, desc: str = None):
+        self.base_task_listener.on_progress(self.sub_task_start + self.sub_task_total, self.base_task_total, desc=desc)

src/vadParallel.py

@@ -170,6 +170,7 @@ class ParallelTranscription(AbstractTranscription):
             results_async = pool.starmap_async(self.transcribe, parameters)
             total_progress = 0
 
+            idx=0
             while not results_async.ready():
                 try:
                     delta = progress_queue.get(timeout=5)  # Set a timeout of 5 seconds
@@ -178,13 +179,14 @@ class ParallelTranscription(AbstractTranscription):
                 
                 total_progress += delta
                 if progress_listener is not None:
-                    progress_listener.on_progress(total_progress, total_duration, desc="Transcribe parallel")
+                    idx+=1
+                    progress_listener.on_progress(total_progress, total_duration, desc=f"Transcribe parallel: {idx}, {total_progress:.2f}/{total_duration}")
 
             results = results_async.get()
 
             # Call the finished callback
             if progress_listener is not None:
-                progress_listener.on_finished()
+                progress_listener.on_finished(desc=f"Transcribe parallel: {idx}, {total_progress:.2f}/{total_duration}.")
 
             for result in results:
                 # Merge the results

src/whisper/fasterWhisperContainer.py

@@ -145,12 +145,12 @@ class FasterWhisperCallback(AbstractWhisperCallback):
         )
 
         segments = []
-
+        
         for segment in segments_generator:
             segments.append(segment)
 
             if progress_listener is not None:
-                progress_listener.on_progress(segment.end, info.duration, "Transcribe")
+                progress_listener.on_progress(segment.end, info.duration, desc=f"Transcribe: {segment_index}")
             if verbose:
                 print("[{}->{}] {}".format(format_timestamp(segment.start, True), format_timestamp(segment.end, True),
                                           segment.text))
@@ -187,7 +187,7 @@ class FasterWhisperCallback(AbstractWhisperCallback):
             self.prompt_strategy.on_segment_finished(segment_index, prompt, detected_language, result)
 
         if progress_listener is not None:
-            progress_listener.on_finished()
+            progress_listener.on_finished(desc=f"Transcribe: {segment_index}.")
         return result
 
     def _split_suppress_tokens(self, suppress_tokens: Union[str, List[int]]):

webui.bat

@@ -1,7 +1,7 @@
 @echo off
 
 :: The source of the webui.bat file is stable-diffusion-webui
-set COMMANDLINE_ARGS=--whisper_implementation faster-whisper --input_audio_max_duration -1 --default_model_name large-v2 --auto_parallel True --output_dir output --vad_max_merge_size 90 --save_downloaded_files --autolaunch
+set COMMANDLINE_ARGS=--whisper_implementation faster-whisper --input_audio_max_duration -1 --default_model_name large-v2 --auto_parallel True --output_dir output --vad_max_merge_size 90 --merge_subtitle_with_sources --autolaunch
 
 if not defined PYTHON (set PYTHON=python)
 if not defined VENV_DIR (set "VENV_DIR=%~dp0%venv")