Let users select start second for the transcription

app.py

@@ -34,7 +34,7 @@ def change_model(model):
     current_model = model
 
 # Credits https://huggingface.co/spaces/rajesh1729/youtube-video-transcription-with-whisper
-def get_audio(url):
+def get_audio(url, start):
     yt = YouTube(url)
     video = yt.streams.filter(only_audio=True).first()
     out_file = video.download(output_path=".")
@@ -45,15 +45,15 @@ def get_audio(url):
     wav_to_cut = AudioSegment.from_file(a)
     # pydub does things in milliseconds
     ten_seconds = 10 * 1000
-    first_10_seconds = wav_to_cut[:ten_seconds]
+    first_10_seconds = wav_to_cut[start:start+ten_seconds]
     os.remove(new_file)
     first_10_seconds.export("final_audio.wav", format="wav")
     return "final_audio.wav"
 
 # Credits https://huggingface.co/spaces/jeffistyping/Youtube-Whisperer
-def populate_metadata(link):
+def populate_metadata(link, start_second):
     yt = YouTube(link)
-    audio = get_audio(link)
+    audio = get_audio(link, start_second)
     return yt.thumbnail_url, yt.title, audio, audio
 
 def inference(yt_audio_path):
@@ -105,7 +105,11 @@ with demo:
         )
         model.change(fn=change_model, inputs=model, outputs=[])
 
-        link = gr.Textbox(label="YouTube Link")
+        with gr.Row():
+            link = gr.Textbox(label="YouTube Link")
+            start_second = gr.Number(label="Select starting point (in seconds) for the transcription",
+                                    value=0,
+                                    precision=0)
         with gr.Row().style(mobile_collapse=False, equal_height=True):
             title = gr.Label(label="Video Title", placeholder="Title")
             img = gr.Image(label="Thumbnail")
@@ -113,7 +117,9 @@ with demo:
             yt_audio = gr.Audio()
             yt_audio_path = gr.Textbox(visible=False)
 
-        link.change(fn=populate_metadata, inputs=link, outputs=[img, title, yt_audio, yt_audio_path])
+        link.change(fn=populate_metadata,
+                    inputs=[link, start_second],
+                    outputs=[img, title, yt_audio, yt_audio_path])
         
         with gr.Row():
             btn = gr.Button("Transcribe music")