new

app.py

@@ -3,7 +3,7 @@ import stable_whisper
 import json
 import torch
 import soundfile as sf
-import io
+from io import BytesIO
 
 # Create a dropdown to select the model
 model_name = st.selectbox("Select a model", ["base", "small", "medium", "large", "large-v2"])
@@ -18,13 +18,16 @@ audiofile = st.file_uploader("Upload an audio file", type=["mp3", "wav"])
 if st.button('Transcribe'):
     if audiofile is not None:
         # Read the audio file into a numpy array
-        audio_data, _ = sf.read(io.BytesIO(audiofile.read()))
-        # Convert the numpy array to a PyTorch tensor and ensure it's float32
+        audio_data, _ = sf.read(BytesIO(audiofile.read()))
+        # Convert the audio data to float
         audio_data = torch.from_numpy(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)
+        if isinstance(result, stable_whisper.WhisperResult):
+            result_json = result.to_dict()  # replace with actual method if exists
+        else:
+            result_json = json.loads(result)
         st.json(result_json)
     else:
         st.write("Please upload an audio file.")