use upload audio instead

app.py

@@ -50,7 +50,7 @@ gradio_ui = gr.Interface(
     article = """
     This demo showcases two pretrained STT models the first model from speechbrain(wave2vec+CTC models)(1,2gb) is 30 times larger compared to the coqui STT (deepspeech model)(45mb).
     """,
-    inputs=[ gr.inputs.Audio(source="microphone", type="file", optional=True, label="Record from microphone")],
+    inputs=[ gr.inputs.Audio(label="Upload Audio File", type="file", optional=False)],
     outputs=[gr.outputs.Textbox(label="Recognized speech from speechbrain model"),
              gr.outputs.Textbox(label="Recognized speech from coqui STT model"),
              gr.outputs.Textbox(label="Recognized speech from NVIDIA conformer ctc large model")]