Update app.py

app.py

@@ -1,7 +1,8 @@
 import streamlit as st
 from transformers import pipeline, AutoModelForSeq2SeqLM, AutoTokenizer
-import coqui_tts
+from TTS.api import TTS
 from tempfile import NamedTemporaryFile
+import os
 
 # Initialize models
 @st.cache_resource
@@ -10,7 +11,7 @@ def load_models():
     summarizer_tokenizer = AutoTokenizer.from_pretrained("google/flan-t5-base")
     question_generator = pipeline("text2text-generation", model="valhalla/t5-small-qg-hl")
     qa_model = pipeline("question-answering", model="deepset/roberta-base-squad2")
-    tts = coqui_tts.TTS("tts_models/en/ljspeech/glow-tts")  # Load voice model
+    tts = TTS(model_name="tts_models/en/ljspeech/tacotron2-DDC", progress_bar=False, gpu=False)
     return summarizer_model, summarizer_tokenizer, question_generator, qa_model, tts
 
 summarizer_model, summarizer_tokenizer, question_generator, qa_model, tts = load_models()
@@ -21,10 +22,11 @@ def summarize_document(text):
     return summarizer_tokenizer.decode(summary_ids[0], skip_special_tokens=True)
 
 def generate_questions(summary):
-    return question_generator(summary)
+    questions = question_generator(summary)
+    return [q["generated_text"] for q in questions]
 
 def generate_audio(text, voice_gender):
-    tts.voice = "male" if voice_gender == "male" else "female"
+    voice = "ljspeech"  # Use default LJSpeech voice for both genders
     audio_file = NamedTemporaryFile(delete=False, suffix=".wav")
     tts.tts_to_file(text=text, file_path=audio_file.name)
     return audio_file.name
@@ -43,10 +45,9 @@ def app():
 
         # Generate dialogue
         questions = generate_questions(summary)
-        dialogue = [{"persona": "male", "text": q["generated_text"]} for q in questions]
-        for idx, item in enumerate(dialogue):
-            if idx % 2 == 1:
-                dialogue[idx]["persona"] = "female"
+        dialogue = []
+        for idx, question in enumerate(questions):
+            dialogue.append({"persona": "male" if idx % 2 == 0 else "female", "text": question})
 
         # Interactive simulation
         st.write("Simulating Conversation:")