Update app.py

app.py

@@ -6,31 +6,48 @@ from pyannote.audio import Pipeline
 
 # instantiate the pipeline
 try:
+    # בדיקה אם ה-token קיים בסביבה והאם הוא לא ריק
+    auth_token = os.environ.get("HUGGINGFACE_READ_TOKEN")
+    if not auth_token:
+        raise ValueError("HUGGINGFACE_READ_TOKEN not found or is empty")
+    
+    print("Token found, attempting to initialize pipeline...")
+
+    # ניסיון לאתחל את ה-Pipeline
     pipeline = Pipeline.from_pretrained(
         "pyannote/speaker-diarization-3.1",
-        use_auth_token=os.environ["HUGGINGFACE_READ_TOKEN"]
+        use_auth_token=auth_token,
+        cache_dir="./cache"  # כדי לנסות להשתמש במטמון
     )
-    # Move the pipeline to the GPU
-    device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
+
+    # העברת ה-Pipeline ל-CPU בלבד, בהתחשב שאתה בגרסה חינמית
+    device = torch.device("cpu")
     pipeline.to(device)
+    print("Pipeline initialized successfully!")
+
 except Exception as e:
     print(f"Error initializing pipeline: {e}")
     pipeline = None
 
-
 def save_audio(audio):
     if pipeline is None:
         return "Error: Pipeline not initialized"
 
-    # Read the uploaded audio file as bytes
-    with open(audio, "rb") as f:
-        audio_data = f.read()
+    try:
+        # קריאה של קובץ האודיו שעלה
+        with open(audio, "rb") as f:
+            audio_data = f.read()
 
-    # Save the uploaded audio file to a temporary location
-    with open("temp.wav", "wb") as f:
-        f.write(audio_data)
+        # שמירה של קובץ האודיו במיקום זמני
+        temp_file = "temp.wav"
+        with open(temp_file, "wb") as f:
+            f.write(audio_data)
 
-    return "temp.wav"
+        print(f"Audio file saved to {temp_file}")
+        return temp_file
+    except Exception as e:
+        print(f"Error saving audio file: {e}")
+        return None
 
 @spaces.GPU(duration=60 * 2)
 def diarize_audio(temp_file, num_speakers, min_speakers, max_speakers):
@@ -38,6 +55,7 @@ def diarize_audio(temp_file, num_speakers, min_speakers, max_speakers):
         return "Error: Pipeline not initialized"
 
     try:
+        # הכנת פרמטרים לפי הקלט של המשתמש
         params = {}
         if num_speakers > 0:
             params["num_speakers"] = num_speakers
@@ -46,19 +64,25 @@ def diarize_audio(temp_file, num_speakers, min_speakers, max_speakers):
         if max_speakers > 0:
             params["max_speakers"] = max_speakers
 
+        print(f"Processing audio file {temp_file} with parameters: {params}")
         diarization = pipeline(temp_file, **params)
+        print("Diarization completed successfully!")
     except Exception as e:
+        print(f"Error processing audio: {e}")
         return f"Error processing audio: {e}"
 
-    # Remove the temporary file
-    os.remove(temp_file)
+    # הסרת הקובץ הזמני לאחר עיבוד
+    try:
+        os.remove(temp_file)
+        print(f"Temporary file {temp_file} removed successfully.")
+    except Exception as e:
+        print(f"Error removing temporary file {temp_file}: {e}")
 
-    # Return the diarization output
     return str(diarization)
-    
+
 def timestamp_to_seconds(timestamp):
     try:
-        # Extracts hour, minute, and second from timestamp and converts to total seconds
+        # המרה של timestamp לשניות
         h, m, s = map(float, timestamp.split(':'))
         return 3600 * h + 60 * m + s
     except ValueError as e:
@@ -66,7 +90,7 @@ def timestamp_to_seconds(timestamp):
         return None
 
 def generate_labels_from_diarization(diarization_output):
-    successful_lines = 0  # Counter for successfully processed lines
+    successful_lines = 0
     labels_path = 'labels.txt'
     try:
         with open(labels_path, 'w') as outfile:
@@ -76,9 +100,11 @@ def generate_labels_from_diarization(diarization_output):
                     parts = line.strip()[1:-1].split(' --> ')
                     start_time = parts[0].strip()
                     end_time = parts[1].split(']')[0].strip()
-                    label = line.split()[-1].strip()  # Extracting the last word as label
+                    label = line.split()[-1].strip()  # לקיחת התווית מהשורה
                     start_seconds = timestamp_to_seconds(start_time)
                     end_seconds = timestamp_to_seconds(end_time)
+                    if start_seconds is None or end_seconds is None:
+                        continue
                     outfile.write(f"{start_seconds}\t{end_seconds}\t{label}\n")
                     successful_lines += 1
                 except Exception as e:
@@ -89,38 +115,29 @@ def generate_labels_from_diarization(diarization_output):
         print(f"Cannot write to file '{labels_path}'. Error: {e}")
         return None
 
-
-
 def process_audio(audio, num_speakers, min_speakers, max_speakers):
-    diarization_result = diarize_audio(save_audio(audio), num_speakers, min_speakers, max_speakers)
+    temp_file = save_audio(audio)
+    if temp_file is None:
+        return "Error saving audio file", None
+
+    diarization_result = diarize_audio(temp_file, num_speakers, min_speakers, max_speakers)
     if diarization_result.startswith("Error"):
-        return diarization_result, None  # Return None for label file link if there's an error
-    else:
-        label_file = generate_labels_from_diarization(diarization_result)
-        return diarization_result, label_file
+        return diarization_result, None
 
+    label_file = generate_labels_from_diarization(diarization_result)
+    return diarization_result, label_file
+
+# ממשק גריידיו
 with gr.Blocks() as demo:
     gr.Markdown("""
     # 🗣️Pyannote Speaker Diarization 3.1🗣️
-
     This model takes an audio file as input and outputs the diarization of the speakers in the audio.
-
     Please upload an audio file and adjust the parameters as needed.
     
-    The maximum length of the audio file that can be processed depends based on the hardware it's running on. If you are on the ZeroGPU HuggingFace Space, it's around **35-40 minutes**.
-
     If you find this space helpful, please ❤ it.
-
-    Join my server for support and open source AI discussion: https://discord.gg/osai
-
-    IF YOU LEAVE ALL THE PARAMETERS BELOW TO 0, IT WILL BE ON AUTO MODE, AUTOMATICALLY DETECTING THE SPEAKERS, ELSE USE THE ONES BELOW FOR MORE COSTUMIZATION & BETTER RESULTS
-
     """)
     audio_input = gr.Audio(type="filepath", label="Upload Audio File")
-    num_speakers_input = gr.Number(label="Number of Speakers", info="Use it only if you know the number of speakers in advance, else leave it to 0 and use the parameters below", value=0)
-    
-    gr.Markdown("Use the following parameters only if you don't know the number of speakers, you can set lower and/or upper bounds on the number of speakers, if instead you know it, leave the following parameters to 0 and use the one above")
-    
+    num_speakers_input = gr.Number(label="Number of Speakers", value=0)
     min_speakers_input = gr.Number(label="Minimum Number of Speakers", value=0)
     max_speakers_input = gr.Number(label="Maximum Number of Speakers", value=0)
     process_button = gr.Button("Process")
@@ -131,5 +148,6 @@ with gr.Blocks() as demo:
         fn=process_audio,
         inputs=[audio_input, num_speakers_input, min_speakers_input, max_speakers_input],
         outputs=[diarization_output, label_file_link]
-)
-demo.launch(share = False)
+    )
+
+demo.launch(share=False)