Create app.py

app.py

@@ -0,0 +1,52 @@
+import streamlit as st
+import torch
+import torchaudio
+import requests
+from io import BytesIO
+
+# Load the Hugging Face model for speech recognition
+model_name = "facebook/wav2vec2-large-xlsr-53"
+model = torch.hub.load('pytorch/fairseq', model_name)
+
+# Create a function to transcribe audio from a URL using the model
+def transcribe_audio(url):
+    # Download the audio file from the URL
+    response = requests.get(url)
+    audio_bytes = BytesIO(response.content)
+
+    # Load the audio file with Torchaudio and apply preprocessing
+    waveform, sample_rate = torchaudio.load(audio_bytes)
+    with torch.no_grad():
+        features = model.feature_extractor(waveform)
+        logits = model.feature_aggregator(features)
+        transcription = model.decoder.decode(logits)
+
+    return transcription[0]['text']
+
+# Define the Streamlit app
+st.title("Speech Recognition with Hugging Face")
+
+# Add a file uploader to allow the user to upload an audio file
+audio_file = st.file_uploader("Upload an audio file", type=["mp3", "wav"])
+
+if audio_file is not None:
+    # Load the audio file with Torchaudio and apply preprocessing
+    waveform, sample_rate = torchaudio.load(audio_file)
+    with torch.no_grad():
+        features = model.feature_extractor(waveform)
+        logits = model.feature_aggregator(features)
+        transcription = model.decoder.decode(logits)
+
+    # Display the transcription
+    st.write("Transcription:")
+    st.write(transcription[0]['text'])
+
+# Add a text input to allow the user to enter a URL of an audio file
+url = st.text_input("Enter the URL of an audio file")
+if url:
+    # Transcribe the audio from the URL using the model
+    transcription = transcribe_audio(url)
+
+    # Display the transcription
+    st.write("Transcription:")
+    st.write(transcription)