Update utils.py

utils.py

@@ -6,15 +6,15 @@ import tiktoken
 import json
 import re
 import tempfile
-import requests
+from gtts import gTTS
 from bs4 import BeautifulSoup
+import requests
 
 groq_client = Groq(api_key=os.environ["GROQ_API_KEY"])
 tokenizer = tiktoken.get_encoding("cl100k_base")
-VOICERSS_API_KEY = os.environ["VOICERSS_API_KEY"]
 
 class DialogueItem(BaseModel):
-    speaker: Literal["John", "Lily"]
+    speaker: Literal["Sarah", "Maria"]
     text: str
 
 class Dialogue(BaseModel):
@@ -57,9 +57,14 @@ def generate_script(system_prompt: str, input_text: str, tone: str, target_lengt
     Generate a complete, well-structured podcast script that:
     1. Starts with a proper introduction
     2. Covers the main points from the input text
-    3. Has a natural flow of conversation between John and Lily
+    3. Has a natural flow of conversation between Sarah (American accent) and Maria (British accent)
     4. Concludes with a summary and sign-off
     5. Fits within the {word_limit} word limit for the target length of {target_length}
+    6. Strongly emphasizes the {tone} tone throughout the conversation
+
+    For a humorous tone, include jokes, puns, and playful banter.
+    For a casual tone, use colloquial language and make it sound like a conversation between college students.
+    For a formal tone, maintain a professional podcast style with well-structured arguments and formal language.
 
     Ensure the script is not abruptly cut off and forms a complete conversation.
     """
@@ -95,14 +100,8 @@ def generate_script(system_prompt: str, input_text: str, tone: str, target_lengt
     return dialogue
 
 def generate_audio(text: str, speaker: str) -> str:
-    voice = "John" if speaker == "John" else "Lily"
-    url = f"http://api.voicerss.org/?key={VOICERSS_API_KEY}&hl=en-us&v={voice}&src={text}"
-    
-    response = requests.get(url)
-    if response.status_code != 200:
-        raise Exception(f"Error generating audio: {response.text}")
-    
-    # Save the raw audio data
+    tld = 'com' if speaker == "Sarah" else 'co.uk'
+    tts = gTTS(text=text, lang='en', tld=tld)
     with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:
-        temp_audio.write(response.content)
+        tts.save(temp_audio.name)
         return temp_audio.name