Update app.py

app.py

@@ -64,21 +64,21 @@ def synthesise(text):
     model_tts = SpeechT5ForTextToSpeech.from_pretrained("crowbarmassage/speecht5_finetuned_voxpopuli_fr")
     vocoder = SpeechT5HifiGan.from_pretrained("microsoft/speecht5_hifigan")
 
-    #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)
 
 
     # Load your dataset from Hugging Face
-    embeddings_dataset = load_dataset("crowbarmassage/MAEmbed")
-    print(embeddings_dataset.features)
-    print(embeddings_dataset[0])
+    #embeddings_dataset = load_dataset("crowbarmassage/MAEmbed")
+    #print(embeddings_dataset.features)
+    #print(embeddings_dataset[0])
 
     # Extract the embedding (assuming it's in a column named 'embedding')
     # Note: Adjust the index [0] if your embedding is at a different position in the dataset.
-    embedding_array = embeddings_dataset[0]['embedding']
+    #embedding_array = embeddings_dataset[0]['embedding']
 
     # Convert the embedding to a PyTorch tensor and add a batch dimension
-    speaker_embeddings = torch.tensor(embedding_array).unsqueeze(0)
+    #speaker_embeddings = torch.tensor(embedding_array).unsqueeze(0)
 
     inputs = processor_tts(text=text, return_tensors="pt")
     speech = model_tts.generate_speech(