update-3

app.py

@@ -1,8 +1,7 @@
 import streamlit as st
 import stable_whisper
 import json
-import soundfile as sf
-import io
+import torch
 
 # Create a dropdown to select the model
 model_name = st.selectbox("Select a model", ["base", "small", "medium", "large", "large-v2"])
@@ -17,12 +16,13 @@ audiofile = st.file_uploader("Upload an audio file", type=["mp3", "wav"])
 if st.button('Transcribe'):
     if audiofile is not None:
         # Read the audio file
-        file_bytes = audiofile.read()
-        audio_data, sample_rate = sf.read(io.BytesIO(file_bytes))
+        audio_data = torch.from_numpy(audiofile.read())
+        # Convert the audio data to float
+        audio_data = audio_data.float()
         # Transcribe the audio file
         result = model.transcribe(audio_data)
         # Convert the result to JSON and display it
         result_json = json.loads(result)
         st.json(result_json)
     else:
-        st.write("Please upload an audio file.")
+        st.write("Please upload an audio file.")

requirements.txt

@@ -1,3 +1,4 @@
 streamlit
 stable-ts
-soundfile
+torch
+numpy