working for snippeds < 1 min

stt_test.py

@@ -4,6 +4,7 @@ import sounddevice as sd
 import numpy as np
 import wave
 import gradio as gr
+import time
 
 # Setup device
 device = "cuda:0" if torch.cuda.is_available() else "cpu"
@@ -32,13 +33,15 @@ pipe = pipeline(
 SAMPLE_RATE = 16000  # Whisper prefers 16 kHz
 FILENAME = "recorded_audio.wav"
 is_recording = False
+start_time = None  # Track the recording start time
 recorded_audio = None
 
 
 def start_recording():
     """Starts recording audio."""
-    global is_recording, recorded_audio
+    global is_recording, recorded_audio, start_time
     is_recording = True
+    start_time = time.time()  # Record the start time
     print("Recording started...")
     recorded_audio = sd.rec(int(SAMPLE_RATE * 60), samplerate=SAMPLE_RATE, channels=1, dtype=np.float32)
     return "Recording... Click 'Stop Recording' to finish."
@@ -46,13 +49,14 @@ def start_recording():
 
 def stop_recording():
     """Stops recording audio and saves it."""
-    global is_recording, recorded_audio
+    global is_recording, recorded_audio, start_time
     if not is_recording:
         return "Not recording!"
     sd.stop()
     is_recording = False
-    print("Recording stopped.")
-    save_audio_to_wav(recorded_audio, FILENAME)
+    elapsed_time = time.time() - start_time  # Calculate elapsed time
+    print(f"Recording stopped. Duration: {elapsed_time:.2f} seconds.")
+    save_audio_to_wav(recorded_audio[:int(SAMPLE_RATE * elapsed_time)], FILENAME)  # Truncate to actual duration
     return "Recording stopped. Click 'Transcribe' to see the result."