change to italian

app.py

@@ -11,15 +11,15 @@ device = "cuda:0" if torch.cuda.is_available() else "cpu"
 asr_pipe = pipeline("automatic-speech-recognition", model="openai/whisper-base", device=device)
 
 # Load text-to-speech checkpoint and speaker embeddings
-processor = SpeechT5Processor.from_pretrained("facebook/mms-tts-spa")
-model = SpeechT5ForTextToSpeech.from_pretrained("facebook/mms-tts-spa").to(device)
+processor = SpeechT5Processor.from_pretrained("juliuserictuliao/speecht5_finetuned_voxpopuli_it")
+model = SpeechT5ForTextToSpeech.from_pretrained("juliuserictuliao/speecht5_finetuned_voxpopuli_it").to(device)
 vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
 
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
 speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0).to(device)
 
 def translate(audio):
-    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "es"})
+    outputs = asr_pipe(audio, max_new_tokens=256, generate_kwargs={"task": "transcribe", "language": "it"})
     return outputs["text"]
 
 def synthesise(text):