Update app.py

app.py

@@ -1,10 +1,8 @@
 from collections import deque
-
 import streamlit as st
 import torch
 from streamlit_player import st_player
 from transformers import AutoModelForCTC, Wav2Vec2Processor
-
 from streaming import ffmpeg_stream
 
 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
@@ -27,11 +25,9 @@ def load_model(model_path="facebook/wav2vec2-large-robust-ft-swbd-300h"):
     processor = Wav2Vec2Processor.from_pretrained(model_path)
     model = AutoModelForCTC.from_pretrained(model_path).to(device)
     return processor, model
-
-
+    
 processor, model = load_model()
 
-
 def stream_text(url, chunk_duration_ms, pad_duration_ms):
     sampling_rate = processor.feature_extractor.sampling_rate
 
@@ -63,16 +59,14 @@ def stream_text(url, chunk_duration_ms, pad_duration_ms):
             else:
                 yield leftover_text
                 leftover_text = ""
-
     yield leftover_text
 
-
 def main():
     state = st.session_state
-    st.header("YouTube Streaming ASR with Robust Wav2Vec2")
+    st.header("Video ASR Streamlit from Youtube Link")
 
     with st.form(key="inputs_form"):
-        state.youtube_url = st.text_input("YouTube URL", "https://www.youtube.com/watch?v=yJmiZ1Mo1cQ")
+        state.youtube_url = st.text_input("YouTube URL", "https://youtu.be/cC1HszE5Hcw?list=PLHgX2IExbFouJoqEr8JMF5MbZSbyC91-L")
         state.chunk_duration_ms = st.slider("Audio chunk duration (ms)", 2000, 10000, 3000, 100)
         state.pad_duration_ms = st.slider("Padding duration (ms)", 100, 5000, 1000, 100)
         submit_button = st.form_submit_button(label="Submit")