Update app.py

app.py

@@ -1,54 +1,53 @@
-import os
-import whisper
-import spacy
-import language_tool_python
 import gradio as gr
 import subprocess
+import os
+from transformers import WhisperProcessor, WhisperForConditionalGeneration
+import language_tool_python
+from pydub import AudioSegment
 from docx import Document
 
 def extract_audio(video_path, audio_path):
-    try:
-        command = f"ffmpeg -i {video_path} -vn -acodec pcm_s16le -ar 16000 -ac 1 {audio_path}"
-        subprocess.run(command, shell=True, check=True)
-        return True
-    except Exception as e:
-        print(f"Error al extraer audio: {e}")
-        return False
+    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
 
 def transcribe_audio(audio_path):
-    model = whisper.load_model("base")
-    result = model.transcribe(audio_path, word_timestamps=True)
-    return result
+    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)
+    transcription = processor.decode(result[0], skip_special_tokens=True)
+    
+    return transcription
 
 def correct_text(text):
     tool = language_tool_python.LanguageTool('es')
     matches = tool.check(text)
     return language_tool_python.utils.correct(text, matches)
 
-def create_word_doc(segments, output_path):
-    doc = Document()
-    for segment in segments:
-        corrected_text = correct_text(segment['text'])
-        doc.add_paragraph(corrected_text)
-    doc.save(output_path)
-    return output_path
-
 def process_video(video_file):
-    audio_path = video_file.replace(".mp4", ".wav")
-    word_output = video_file.replace(".mp4", "_transcription.docx")
+    video_path = video_file.name
+    audio_path = os.path.splitext(video_path)[0] + '.wav'
+    
+    extract_audio(video_path, audio_path)
+    transcribed_text = transcribe_audio(audio_path)
+    corrected_text = correct_text(transcribed_text)
+    
+    doc = Document()
+    doc.add_paragraph(corrected_text)
+    doc_path = "transcription.docx"
+    doc.save(doc_path)
     
-    if extract_audio(video_file, audio_path):
-        result = transcribe_audio(audio_path)
-        segments = result['segments']
-        doc_path = create_word_doc(segments, word_output)
-        return "Transcripción completada.", doc_path
-    else:
-        return "Error al procesar el archivo.", None
+    return corrected_text, doc_path
 
 demo = gr.Interface(
     fn=process_video,
     inputs=gr.File(label="Sube un archivo de video"),
-    outputs=["text", gr.File(label="Descargar transcripción")]
+    outputs=[
+        gr.Textbox(label="Texto transcrito y corregido"),
+        gr.File(label="Descargar transcripción Word")
+    ]
 )
 
 demo.launch()