Update app.py

app.py

@@ -31,6 +31,7 @@ from fpdf import FPDF
 from pathlib import Path
 
 import numpy as np
+import librosa
 import torch
 from transformers import pipeline 
 
@@ -54,6 +55,61 @@ CACHE_EXAMPLES = torch.device('cuda') and os.getenv("CACHE_EXAMPLES", "0") == "1
 device = torch.device('cuda')
 #device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")
 
+def transcribe(file_upload, progress=gr.Progress(track_tqdm=True)):
+
+    file = file_upload # microphone if microphone is not None else 
+    start_time = time.time()
+    
+    # -- ex subrosa
+    audio, sr = librosa.load(file_upload, sr=None)
+    duration = librosa.get_duration(y=audio, sr=sr)
+    
+    # -- asr pipeline
+    with torch.no_grad():
+        pipe = pipeline("automatic-speech-recognition", model="NbAiLab/nb-whisper-large", chunk_length_s=30, device=device)
+    
+    # -- process audio in chunks of 30 seconds
+    chunk_size = sr * 30  # 30 seconds
+    text = ""
+    for start in range(0, len(audio), chunk_size):
+        end = min(start + chunk_size, len(audio))
+        chunk = audio[start:end]
+        
+        # -- convert audio chunk to format for pipeline
+        chunk_file = "temp_chunk.wav"
+        librosa.output.write_wav(chunk_file, chunk, sr)
+        
+        # -- chnk ad transcriptrauma
+        chunk_text = pipe(chunk_file)["text"]
+        text += chunk_text + " "
+
+    end_time = time.time()
+    output_time = end_time - start_time
+    
+    # --word count
+    word_count = len(text.split())
+
+    # --memory metrics
+    memory = psutil.virtual_memory()
+    
+    # --cpu metric
+    #cpu_usage = psutil.cpu_percent(interval=1)
+
+    # --gpu metric
+    #gpu_utilization, gpu_memory = GPUInfo.gpu_usage()
+    
+    # --system info string
+    system_info = f"""
+    #Processing time: {output_time:.2f} seconds.
+    #Number of words: {word_count}
+    GPU Memory: {gpu_memory}%
+    GPU Utilization: {gpu_utilization}%
+    """
+    
+    return text.strip(), system_info
+
+
+"""
 #@spaces.GPU
 def transcribe(file_upload, progress=gr.Progress(track_tqdm=True)): # microphone
     
@@ -86,9 +142,9 @@ def transcribe(file_upload, progress=gr.Progress(track_tqdm=True)): # microphone
     
     # --system info string
     system_info = f"""
-    Processing time: {output_time:.2f} seconds.
-    Number of words: {word_count}
-    GPU Memory: {gpu_memory}"""
+    #Processing time: {output_time:.2f} seconds.
+    #Number of words: {word_count}
+    #GPU Memory: {gpu_memory}"""
  
     #--------------____________________________________________--------------"
     
@@ -96,8 +152,9 @@ def transcribe(file_upload, progress=gr.Progress(track_tqdm=True)): # microphone
     #Memory used: {memory.percent}%
     #GPU Utilization: {gpu_utilization}%
     
-    return text, system_info
+    #return text, system_info
 
+"""
 
 ###############################################################################
 # Interface.