Update app.py

app.py

@@ -10,39 +10,27 @@ device = "cuda:0" if torch.cuda.is_available() else "cpu"
 model = CombinedModel("facebook/mms-1b-all", "Sunbird/sunbird-mul-en-mbart-merged", device=device)
 
 def transcribe(audio_file_mic=None, audio_file_upload=None):
-    if audio_file_mic:
-        audio_file = audio_file_mic
-    elif audio_file_upload:
-        audio_file = audio_file_upload
-    else:
-        return "Please upload an audio file or record one"
-
-    # Make sure audio is 16kHz
-    speech, sample_rate = librosa.load(audio_file)
-    if sample_rate != 16000:
-        speech = librosa.resample(speech, orig_sr=sample_rate, target_sr=16000)
-    speech = torch.tensor([speech])
-
-    with torch.no_grad():
-        transcription, translation = model({"audio":speech})
-
-    return transcription, translation[0]
+    # Your transcribe function code here
+    # ...
 
 description = '''Luganda to English Speech Translation'''
 
 # Define example audio files
 example_audio_files = [
-    "audio/luganda.mp3"  # Replace with the path to your first example audio file
+    "audio/luganda.mp3",  # Replace with the path to your first example audio file
+    #"example_audio_files/example2.wav",  # Replace with the path to your second example audio file
 ]
 
 # Generate example inputs and outputs
 examples = []
 for audio_file_path in example_audio_files:
     transcription, translation = transcribe(audio_file_upload=audio_file_path)
-    examples.append({
-        "input": audio_file_path,
-        "output": [transcription, translation]
-    })
+    examples.append([
+        audio_file_path,  # First element corresponds to the first input component (audio_file_upload)
+        None,  # Second element corresponds to the second input component (audio_file_mic). Set to None for this example.
+        transcription,
+        translation
+    ])
 
 iface = gr.Interface(
     fn=transcribe,