Update app.py

app.py

@@ -1,39 +1,57 @@
 import gradio as gr
 import subprocess
 import os
+import librosa
 from transformers import WhisperProcessor, WhisperForConditionalGeneration
 import language_tool_python
 from pydub import AudioSegment
 from docx import Document
 
+# Función para extraer audio de video
 def extract_audio(video_path, audio_path):
     command = f"ffmpeg -i '{video_path}' -ar 16000 -ac 1 -c:a pcm_s16le '{audio_path}' -y"
     subprocess.run(command, shell=True, check=True)
     return audio_path
 
+# Función para transcribir el audio usando Whisper
 def transcribe_audio(audio_path):
+    # Cargar el procesador y modelo de Whisper
     processor = WhisperProcessor.from_pretrained("openai/whisper-base")
     model = WhisperForConditionalGeneration.from_pretrained("openai/whisper-base")
     
-    audio_input = processor(audio_path, return_tensors="pt", sampling_rate=16000)
-    result = model.generate(**audio_input)
+    # Cargar el archivo de audio usando librosa
+    audio_input, _ = librosa.load(audio_path, sr=16000)
+    
+    # Preprocesar el audio para el modelo
+    inputs = processor(audio_input, return_tensors="pt", sampling_rate=16000)
+    
+    # Realizar la transcripción
+    result = model.generate(**inputs)
     transcription = processor.decode(result[0], skip_special_tokens=True)
     
     return transcription
 
+# Función para corregir el texto transcrito con LanguageTool
 def correct_text(text):
     tool = language_tool_python.LanguageTool('es')
     matches = tool.check(text)
     return language_tool_python.utils.correct(text, matches)
 
+# Función principal que procesa el video
 def process_video(video_file):
     video_path = video_file.name
     audio_path = os.path.splitext(video_path)[0] + '.wav'
     
+    # Extraer el audio del video
     extract_audio(video_path, audio_path)
+    
+    # Transcribir el audio
     transcribed_text = transcribe_audio(audio_path)
+    
+    # Corregir la transcripción
     corrected_text = correct_text(transcribed_text)
     
+    # Crear un documento Word con la transcripción corregida
     doc = Document()
     doc.add_paragraph(corrected_text)
     doc_path = "transcription.docx"
@@ -41,6 +59,7 @@ def process_video(video_file):
     
     return corrected_text, doc_path
 
+# Interfaz de Gradio
 demo = gr.Interface(
     fn=process_video,
     inputs=gr.File(label="Sube un archivo de video"),