Update utils.py

utils.py

@@ -8,16 +8,13 @@ import re
 import tempfile
 import requests
 from bs4 import BeautifulSoup
-from TTS.api import TTS
-import torch
+import subprocess
+import pyttsx3
+from pydub import AudioSegment
 
 groq_client = Groq(api_key=os.environ["GROQ_API_KEY"])
 tokenizer = tiktoken.get_encoding("cl100k_base")
 
-# Initialize TTS models
-tts_maria = TTS("tts_models/en/ljspeech/tacotron2-DDC")
-tts_sarah = TTS("tts_models/en/ljspeech/glow-tts")
-
 class DialogueItem(BaseModel):
     speaker: Literal["Maria", "Sarah"]
     text: str
@@ -99,10 +96,23 @@ def generate_script(system_prompt: str, input_text: str, tone: str, target_lengt
     
     return dialogue
 
-def generate_audio(text: str, speaker: str) -> str:
+def generate_audio_espeak(text: str, speaker: str) -> str:
+    voice = "en-us+f3" if speaker == "Maria" else "en-gb+f3"
+    with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
+        subprocess.call(['espeak-ng', '-v', voice, '-w', temp_audio.name, text])
+        return temp_audio.name
+
+def generate_audio_pyttsx3(text: str, speaker: str) -> str:
+    engine = pyttsx3.init()
+    voices = engine.getProperty('voices')
+    engine.setProperty('voice', voices[1].id if speaker == "Maria" else voices[0].id)
     with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as temp_audio:
-        if speaker == "Maria":
-            tts_maria.tts_to_file(text=text, file_path=temp_audio.name)
-        else:  # Sarah
-            tts_sarah.tts_to_file(text=text, file_path=temp_audio.name)
-        return temp_audio.name
+        engine.save_to_file(text, temp_audio.name)
+        engine.runAndWait()
+        return temp_audio.name
+
+def generate_audio(text: str, speaker: str) -> str:
+    try:
+        return generate_audio_espeak(text, speaker)
+    except Exception:
+        return generate_audio_pyttsx3(text, speaker)