Update backend.py

backend.py

@@ -30,8 +30,5 @@ def text_to_speech(text):
     inputs = tokenizer_tts(text, return_tensors="pt").to(model_tts.device)
     with torch.no_grad():
         audio = model_tts(**inputs).waveform.squeeze().cpu().numpy()
-    buffer = io.BytesIO()
     sampling_rate = model_tts.config.sampling_rate
-    sf.write(buffer, audio, sampling_rate, format="WAV")
-    buffer.seek(0)
-    return sampling_rate, buffer
+    return sampling_rate, audio