adds get duration endpoint

app.py

@@ -1,10 +1,11 @@
-from datetime import datetime, timedelta
 import os
 
 import gradio as gr
 import torch
 import nemo.collections.asr as nemo_asr
 import wandb
+from pydub.utils import mediainfo
+
 
 MODEL_HISTORY_DAYS = 180
 WANDB_ENTITY = os.environ.get("WANDB_ENTITY", "tarteel")
@@ -43,6 +44,12 @@ def transcribe(audio_file):
     return transcription_file
 
 
+def get_duration_ms(audio_file):
+    duration = mediainfo(audio_file)["duration"]
+    duration_ms = int(duration) * 1000
+    return duration_ms
+
+
 demo = gr.Blocks()
 
 with demo:
@@ -66,4 +73,16 @@ with demo:
         api_name="transcribe",
     )
 
+    b2 = gr.Button("Get Duration")
+
+    with gr.Row():
+        duration = gr.TextArea(label="Duration")
+
+    b2.click(
+        get_duration_ms,
+        inputs=[audio_file],
+        outputs=[duration],
+        api_name="get_duration_ms",
+    )
+
 demo.launch()

requirements.txt

@@ -2,3 +2,4 @@ torchaudio
 nemo_toolkit[asr]==1.19.0
 hydra-core
 wandb
+pydub