transcribe files as well

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=False, label="Record from microphone")],
+    inputs=[gr.inputs.Audio(label="Upload Audio File", type="file", optional=True), gr.inputs.Audio(source="microphone", type="file", optional=False, label="Record from microphone")],
     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 transduver large model")]

requirements.txt

@@ -10,4 +10,5 @@ ffmpeg-python
 soundfile==0.10.3.post1
 wget
 aiofiles
+pydub
 git+https://github.com/NVIDIA/NeMo.git@r1.11.0#egg=nemo_toolkit[all]