Update app.py

app.py

@@ -3,10 +3,10 @@ import numpy as np
 import torch
 from datasets import load_dataset
 
-#from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline
-from transformers import pipeline
-from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
-from transformers import BarkModel, BarkProcessor
+from transformers import SpeechT5ForTextToSpeech, SpeechT5HifiGan, SpeechT5Processor, pipeline
+#from transformers import pipeline
+#from transformers import Wav2Vec2Processor, Wav2Vec2ForCTC
+#from transformers import BarkModel, BarkProcessor
 
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
 
@@ -14,12 +14,12 @@ device = "cuda:0" if torch.cuda.is_available() else "cpu"
 asr_pipe = pipeline("automatic-speech-recognition", model="facebook/wav2vec2-base-100h", device=device)
 
 # load text-to-speech checkpoint and speaker embeddings
-#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)
+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 = BarkModel.from_pretrained("suno/bark-small")
-processor = BarkProcessor.from_pretrained("suno/bark-small")
+#model = BarkModel.from_pretrained("suno/bark-small")
+#processor = BarkProcessor.from_pretrained("suno/bark-small")
 
 embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
 speaker_embeddings = torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0)
@@ -32,8 +32,8 @@ def translate(audio):
 
 def synthesise(text):
     inputs = processor(text=text, return_tensors="pt")
-    #speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
-    speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device))
+    speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device), vocoder=vocoder)
+    #speech = model.generate_speech(inputs["input_ids"].to(device), speaker_embeddings.to(device))
     return speech.cpu()
 
 
@@ -46,8 +46,7 @@ def speech_to_speech_translation(audio):
 
 title = "Cascaded STST"
 description = """
-Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in English. Demo uses OpenAI's [Whisper Base](https://huggingface.co/openai/whisper-base) model for speech translation, and Microsoft's
-[SpeechT5 TTS](https://huggingface.co/microsoft/speecht5_tts) model for text-to-speech:
+Demo for cascaded speech-to-speech translation (STST), mapping from source speech in any language to target speech in English. Demo uses facebook/wav2vec2-base-100h model for speech translation, and microsoft/speecht5_tts model for text-to-speech:
 
 ![Cascaded STST](https://huggingface.co/datasets/huggingface-course/audio-course-images/resolve/main/s2st_cascaded.png "Diagram of cascaded speech to speech translation")
 """