Update app.py

app.py

@@ -12,17 +12,17 @@ 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("microsoft/speecht5_tts")
+processor = SpeechT5Processor.from_pretrained("microsoft/speecht5_tts")
 
-# model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts").to(device)
-# vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
+model = SpeechT5ForTextToSpeech.from_pretrained("microsoft/speecht5_tts").to(device)
+vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan").to(device)
 
 
 model = VitsModel.from_pretrained("Matthijs/mms-tts-deu").to(device)
 tokenizer = VitsTokenizer.from_pretrained("Matthijs/mms-tts-deu")
 
-# embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
-# speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
+embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
+speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
 
 
 def translate(audio):
@@ -37,10 +37,11 @@ def synthesise(text):
 
     with torch.no_grad():
         outputs = model(input_ids)
-    return outputs['audio'].cpu()
-    # inputs = processor(text=text, return_tensors="pt")
-    # speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
-    # return speech.cpu()
+    print(outputs)
+    inputs = processor(text=text, return_tensors="pt")
+    speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
+    print(speech)
+    return speech.cpu()
 
 
 def speech_to_speech_translation(audio):