Update app.py

app.py

@@ -4,8 +4,13 @@ import torch
 from diffusers import DiffusionPipeline
 from datasets import load_dataset
 
+#Headline Generator Pipeline
 headline_gen = pipeline("text2text-generation", model="Michau/t5-base-en-generate-headline", tokenizer="t5-base")
+
+#Arabic to English translator Pipeline
 ar_to_en_translator = pipeline("translation", model="Helsinki-NLP/opus-mt-ar-en")
+
+#English to Arabic translator Pipeline
 en_to_ar_translator = pipeline("translation", model="Helsinki-NLP/opus-mt-en-ar")
 
 # Arabic: text-to-speech
@@ -16,27 +21,31 @@ synthesiser_english = pipeline("text-to-speech", model="microsoft/speecht5_tts")
 embeddings_dataset_english = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
 speaker_embedding_english = torch.tensor(embeddings_dataset_english[7306]["xvector"]).unsqueeze(0)
 
+#main function
 def generate_headline(selected_language, text):
     if selected_language == "Arabic":
-        translated_text = translate_ar_to_en(text)  # Translate Arabic to English
-        english_headline = generate_headline_english(translated_text)  # Generate headline in English
-        arabic_headline = translate_en_to_ar(english_headline)  # Translate headline back to Arabic
-        sampling_rate, audio_data = text_to_speech_arabic(arabic_headline)
+        translated_text = translate_ar_to_en(text)  #Translate Arabic text to English (so the headlines pipeline will understand the input)
+        english_headline = generate_headline_english(translated_text)  #Generate headline in English based on the translated text
+        arabic_headline = translate_en_to_ar(english_headline)  # Translate headline to Arabic (output will be arabic)
+        sampling_rate, audio_data = text_to_speech_arabic(arabic_headline) #Arabic headline will be passed to the text-to-speech pipeline (arabic)
         return arabic_headline, (sampling_rate, audio_data)
 
     elif selected_language == "English":
-        english_headline = generate_headline_english(text)
-        sampling_rate, audio_data = text_to_speech_english(english_headline)
+        english_headline = generate_headline_english(text) #Generate headline in English based on the text
+        sampling_rate, audio_data = text_to_speech_english(english_headline) #English headline will be passed to the text-to-speech pipeline (english)
         return english_headline, (sampling_rate, audio_data)
 
+#function to translate Arabic Text to English
 def translate_ar_to_en(text):
   var_ar_to_en = ar_to_en_translator(text)[0]['translation_text']
   return var_ar_to_en
 
+#function to translate English Headline to Arabic
 def translate_en_to_ar(text):
   var_en_to_ar = en_to_ar_translator(text)[0]['translation_text']
   return var_en_to_ar
 
+#function to generate headline in english
 def generate_headline_english(text):
     result1 = headline_gen(text, max_length=100, truncation=True)
     result2 = result1[0]['generated_text']
@@ -111,5 +120,4 @@ interface = gr.Interface(
     css=custom_css  
 )
 
-interface.launch()
-
+interface.launch()