update

app.py

@@ -38,10 +38,13 @@ coquiTTS = CoquiTTS()
 
   
 # Driver function
-def driver_fun(audio) : 
-    
-  translation, lang = whisper_stt(audio)  
+def driver_fun(audio, text) : 
   
+  if text == 'dummy':
+    translation, lang = whisper_stt(audio)  
+  else:
+    translation = text 
+    
   random_val = random.randrange(0,231657)
   if random_val < 226657:
     lower_limit = random_val
@@ -51,7 +54,7 @@ def driver_fun(audio) :
     upper_limit = random_val 
   print(f"lower_limit : upper_limit = {lower_limit} : {upper_limit}")  
   dataset_subset = filtered_dataset['Joke'][lower_limit : upper_limit]
-  data = query({"inputs": {"source_sentence": "That is a happy person","sentences": dataset_subset} } )
+  data = query({"inputs": {"source_sentence": translation ,"sentences": dataset_subset} } ) #"That is a happy person"
   if 'error' in data:
     print(f"Error is : {data}")
     return 'Error in model inference - Run Again Please', 'Error in model inference - Run Again Please', None
@@ -112,10 +115,11 @@ with demo:
       out_transcript = gr.Textbox(label= 'Transcript of your Audio using OpenAI Whisper')
       
     with gr.Column():
+      in_text = gr.Textbox(label='Or enter any text here..', value='dummy')
       out_audio = gr.Audio(label='Audio response form CoquiTTS')  
       out_generated_joke = gr.Textbox(label= 'Joke returned! ')
       
-      b1.click(driver_fun,inputs=[in_audio], outputs=[out_transcript, out_generated_joke, out_audio]) #out_translation_en, out_generated_text,out_generated_text_en, 
+      b1.click(driver_fun,inputs=[in_audio, in_text], outputs=[out_transcript, out_generated_joke, out_audio]) #out_translation_en, out_generated_text,out_generated_text_en, 
   with gr.Row():
     gr.Markdown(
         """Model pipeline consisting of - <br>- [**Whisper**](https://github.com/openai/whisper) for Speech-to-text, <br>- [**CoquiTTS**](https://huggingface.co/coqui)  for Text-To-Speech.<br>- [Sentence Transformers](https://huggingface.co/models?library=sentence-transformers&sort=downloads)<br>- Front end is built using [**Gradio Block API**](https://gradio.app/docs/#blocks).<br><be>If you want to reuse the App, simply click on the small cross button in the top right corner of your voice record panel, and then press record again! <br><br> Few Caveats:<br>1. Please note that sometimes the joke might be NSFW. Although, I have tried putting in filters to not have that experience, but they seem non-exhaustive.<br>2. Sometimes the joke might not match your theme, please bear with the limited capabilities of free open-source ML prototypes.<br>3. Much like real life, sometimes the joke might just not land, haha!<br>4. If you see the message 'Error in model inference - Run Again Please', just press the button again every time!