support manual multilang detection

app.py

@@ -4,10 +4,10 @@ from pytube import YouTube
 
 loaded_model = whisper.load_model("base")
 current_size = 'base'
-def inference(link):
+def inference(link, lang):
   yt = YouTube(link)
   path = yt.streams.filter(only_audio=True)[0].download(filename="audio.mp4")
-  options = whisper.DecodingOptions(without_timestamps=True)
+  options = whisper.DecodingOptions(without_timestamps=True, language=lang)
   results = loaded_model.transcribe(path)
   return results['text']
 
@@ -42,6 +42,8 @@ with block:
         with gr.Box():
           sz = gr.Dropdown(label="Model Size", choices=['base','small', 'medium', 'large'], value='base')
           
+          lang = gr.Textbox(label="Languange")
+          
           link = gr.Textbox(label="YouTube Link")
           
           with gr.Row().style(mobile_collapse=False, equal_height=True):
@@ -55,7 +57,7 @@ with block:
               btn = gr.Button("Transcribe")       
           
           # Events
-          btn.click(inference, inputs=[link], outputs=[text])
+          btn.click(inference, inputs=[link, lang], outputs=[text])
           link.change(populate_metadata, inputs=[link], outputs=[img, title])
           sz.change(change_model, inputs=[sz], outputs=[])