load a lighter model and check

app.py

@@ -5,7 +5,7 @@ import torch
 # Load lightweight models
 transcriber = pipeline("automatic-speech-recognition", model="openai/whisper-tiny")
 translator = pipeline("translation", model="facebook/m2m100_418M")
-tts = pipeline("text-to-speech", model="espnet/kan-bayashi_ljspeech_vits", device=torch.device('cpu'))
+tts = pipeline("text-to-speech", model="microsoft/speecht5_tts", device=torch.device('cpu'))
 
 # Function to handle transcription, translation, and TTS
 def process_audio(audio, target_lang):