Update app.py

app.py

@@ -1,9 +1,9 @@
 import streamlit as st
-from transformers import pipeline
-import soundfile as sf
+from transformers import pipeline, SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan
 import torch
-from transformers import SpeechT5Processor, SpeechT5ForTextToSpeech, SpeechT5HifiGan, TextToSpeechPipeline
+from transformers import TextToSpeechPipeline
 from datasets import load_dataset
+import io
 
 # Initialize translation models
 translator_urdu = pipeline("translation_en_to_ur", model="Helsinki-NLP/opus-mt-en-ur")
@@ -34,9 +34,12 @@ def synthesize_speech(text, target_lang):
     speaker_embedding = torch.tensor(xvectors[0]["xvector"]).unsqueeze(0)
     speech = tts_model.generate_speech(inputs["input_ids"], speaker_embedding, vocoder=vocoder)
     
-    audio_path = "output.wav"
-    sf.write(audio_path, speech.numpy(), 16000)
-    return audio_path, None
+    # Convert the speech to an in-memory WAV file using io.BytesIO
+    audio_io = io.BytesIO()
+    audio_io.write(speech.numpy().tobytes())
+    audio_io.seek(0)
+    
+    return audio_io, None
 
 # Streamlit UI
 st.title("Language Translator")
@@ -63,4 +66,4 @@ if st.button("Translate"):
         if tts_error:
             st.error(tts_error)
         else:
-            st.audio(audio_file)
+            st.audio(audio_file, format='audio/wav')