Update app.py

app.py

@@ -1,5 +1,6 @@
 from __future__ import unicode_literals
 import youtube_dl
+import yt_dlp
 from pydub import AudioSegment
 from pyannote.audio import Pipeline
 import re
@@ -24,12 +25,6 @@ import contextlib
 from sklearn.cluster import AgglomerativeClustering
 import numpy as np
 
-model = whisper.load_model("medium")
-embedding_model = PretrainedSpeakerEmbedding(
-    "speechbrain/spkrec-ecapa-voxceleb",
-    device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
-)
-
 
 __FILES = set()
       
@@ -49,7 +44,7 @@ def RemoveAllFiles():
     
 def Transcribe_V1(NumberOfSpeakers, SpeakerNames="", audio="temp_audio.wav"):
     SPEAKER_DICT = {}
-    SPEAKERS = [speaker.strip() for speaker in SpeakerNames.split(',')]
+    SPEAKERS = [speaker.strip() for speaker in SpeakerNames.split(',') if len(speaker)]
     
     def GetSpeaker(sp):
         speaker = sp
@@ -109,7 +104,7 @@ def Transcribe_V1(NumberOfSpeakers, SpeakerNames="", audio="temp_audio.wav"):
         return f"dz_{audio}.wav", dzList, segments
     
     def transcribe(dz_audio):
-        model = whisper.load_model("large")
+        model = whisper.load_model("medium")
         result = model.transcribe(dz_audio)
         # for _ in result['segments']:
         #     print(_['start'], _['end'], _['text'])
@@ -144,8 +139,13 @@ def Transcribe_V1(NumberOfSpeakers, SpeakerNames="", audio="temp_audio.wav"):
 
 
 def Transcribe_V2(num_speakers, speaker_names, audio="temp_audio.wav"):
+    model = whisper.load_model("medium")
+    embedding_model = PretrainedSpeakerEmbedding(
+        "speechbrain/spkrec-ecapa-voxceleb",
+        device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
+    )
     SPEAKER_DICT = {}
-    SPEAKERS = [speaker.strip() for speaker in speaker_names.split(',')]
+    SPEAKERS = [speaker.strip() for speaker in speaker_names.split(',') if len(speaker)]
     def GetSpeaker(sp):
         speaker = sp
         if sp not in list(SPEAKER_DICT.keys()):
@@ -223,6 +223,7 @@ def Transcribe_V2(num_speakers, speaker_names, audio="temp_audio.wav"):
     # return output
 
 def AudioTranscribe(NumberOfSpeakers=None, SpeakerNames="", audio="", retries=5):
+    print(f"{NumberOfSpeakers}, {SpeakerNames}, {retries}")
     if retries:
         # subprocess.call(['ffmpeg', '-i', audio,'temp_audio.wav'])
         try:
@@ -268,7 +269,7 @@ def YoutubeTranscribe(NumberOfSpeakers=None, SpeakerNames="", URL="", retries =
                 }],
             }
             try:
-              with youtube_dl.YoutubeDL(ydl_opts) as ydl:
+              with yt_dlp.YoutubeDL(ydl_opts) as ydl:
                   ydl.download([URL])
             except:
                 return YoutubeTranscribe(NumberOfSpeakers, SpeakerNames, URL, retries-1)
@@ -295,5 +296,43 @@ at = gr.Interface(
     outputs=[gr.Textbox(label="Transcribed Text", lines=15), gr.JSON(label="Transcribed JSON")]
 )
 
-demo = gr.TabbedInterface([ut, vt, at], ["Youtube URL", "Video", "Audio"])
-demo.launch()
+# demo = gr.TabbedInterface([ut, vt, at], ["Youtube URL", "Video", "Audio"])
+# demo.launch()
+with gr.Blocks() as yav_ui:
+    with gr.Tab("Input"):
+        with gr.Tab("Youtube", id=1):
+            yinput_nos = gr.Number(label="Number of Speakers", placeholder="2")
+            yinput_sn = gr.Textbox(label="Name of the Speakers (ordered by the time they speak and separated by comma)", placeholder="If Speaker 1 is first to speak followed by Speaker 2 then -> Speaker 1, Speaker 2")
+            yinput = gr.Textbox(label="Youtube Link", placeholder="https://www.youtube.com/watch?v=GECcjrYHH8w")
+            ybutton_transcribe = gr.Button("Transcribe")
+        with gr.Tab("Video", id=2):
+            vinput_nos = gr.Number(label="Number of Speakers", placeholder="2")
+            vinput_sn = gr.Textbox(label="Name of the Speakers (ordered by the time they speak and separated by comma)", placeholder="If Speaker 1 is first to speak followed by Speaker 2 then -> Speaker 1, Speaker 2")
+            vinput = gr.Video(label="Video")
+            vbutton_transcribe = gr.Button("Transcribe")
+        with gr.Tab("Audio", id=3):
+            ainput_nos = gr.Number(label="Number of Speakers", placeholder="2")
+            ainput_sn = gr.Textbox(label="Name of the Speakers (ordered by the time they speak and separated by comma)", placeholder="If Speaker 1 is first to speak followed by Speaker 2 then -> Speaker 1, Speaker 2")
+            ainput = gr.Audio(label="Audio")
+            abutton_transcribe = gr.Button("Transcribe")
+    with gr.Tab("Output"):
+        with gr.Tab("Text"):
+            output_textbox = gr.Textbox(label="Transcribed Text", lines=15)
+        with gr.Tab("JSON"):
+            output_json = gr.JSON(label="Transcribed JSON")
+    ybutton_transcribe.click(
+                fn=YoutubeTranscribe,
+                inputs=[yinput_nos,yinput_sn,yinput],
+                outputs=[output_textbox,output_json]
+            )
+    abutton_transcribe.click(
+                fn=AudioTranscribe,
+                inputs=[ainput_nos,ainput_sn,ainput],
+                outputs=[output_textbox,output_json]
+            )
+    vbutton_transcribe.click(
+                fn=VideoTranscribe,
+                inputs=[vinput_nos,vinput_sn,vinput],
+                outputs=[output_textbox,output_json]
+            )
+yav_ui.launch(debug=True)