Update app.py

app.py

@@ -2,20 +2,21 @@ import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer
 
 # Load the model and tokenizer
-model_name = "suno/bark-small"
+model_name = "facebook/wav2vec2-large-xlsr-53"
 tokenizer = AutoTokenizer.from_pretrained(model_name)
 model = AutoModelForCausalLM.from_pretrained(model_name)
 
 # Define the Gradio interface
 def text_to_speech(text):
     # Tokenize the input text
-    inputs = tokenizer.encode(text, return_tensors="pt")
+    inputs = tokenizer(text, return_tensors="pt", padding=True)
 
     # Generate speech from the input text using the loaded model
-    outputs = model.generate(inputs)
-
+    with torch.no_grad():
+        outputs = model.generate(**inputs)
+    
     # Convert the generated speech tensor to audio format
-    speech = gradio.inputs.Audio(outputs)
+    speech = gradio.inputs.Audio(outputs[0].numpy().tolist(), type='torch')
 
     return speech