Update app.py

app.py

@@ -11,52 +11,63 @@ def get_thumbnail(image_path, width):
     image.thumbnail((width, width))
     return image
 
-model = GPT2LMHeadModel.from_pretrained("salomonsky/deepSP")
+generator = pipeline('text-generation', model="checkpoints")
+tokenizer = GPT2Tokenizer.from_pretrained('checkpoints')
 os.environ["TOKENIZERS_PARALLELISM"] = "true"
-generator = pipeline('text-generation', model="salomonsky/deepSP")
-tokenizer = GPT2Tokenizer.from_pretrained('salomonsky/deepSP')
 
 def generate_output(name, date_of_birth, image):
-    prompt = f"Tu frase mágica de hoy {name} es:"
+    prompt = f"Tu carta astral de hoy {name} es:"
     input_tokens = tokenizer.encode(prompt, add_special_tokens=False)
     input_text = tokenizer.decode(input_tokens)
     gpt2_output = generator(input_text, max_length=120, do_sample=True, temperature=0.9)
     generated_text = gpt2_output[0]['generated_text']
     generated_text = generated_text.replace(input_text, "").strip()
+    if len(gpt2_output) == 0 or 'generated_text' not in gpt2_output[0]:
+        return None, "No se pudo generar el texto."
 
-    tts = gTTS(generated_text, lang='es')
-    temp_audio_path = "temp_audio.mp3"
-    tts.save(temp_audio_path)
-    audio_path = "audio.wav"
-    audio = AudioSegment.from_mp3(temp_audio_path)
-    audio.export(audio_path, format="wav")
+    try:
+        tts = gTTS(generated_text, lang='es')
+        temp_audio_path = "temp_audio.mp3"
+        tts.save(temp_audio_path)
+        audio_path = "audio.wav"
+        audio = AudioSegment.from_mp3(temp_audio_path)
+        audio.export(audio_path, format="wav")
+        print("Archivo de audio generado:", audio_path)
+    except Exception as e:
+        return None, f"No se pudo generar el audio: {str(e)}"   
 
     command = f"python3 inference.py --checkpoint_path checkpoints/wav2lip_gan.pth --face {image} --audio audio.wav --outfile video.mp4 --nosmooth"
     process = subprocess.run(command, shell=True, stdout=subprocess.PIPE, stderr=subprocess.PIPE)
     if process.returncode != 0:
         error_message = process.stderr.decode("utf-8")
-        return None, error_message
+        return None, f"No se pudo generar el video: {error_message}"
 
     output_video_path = "video.mp4"
     os.remove(temp_audio_path)
 
     if os.path.isfile(output_video_path):
         return output_video_path, None
-
     return None, "No se pudo generar el video"
 
 choices = ["1.jpg", "2.jpg", "3.jpg", "4.jpg", "5.jpg", "6.jpg"]
 thumbnail_width = "50px"
 
+
+def error_message_fn(error_message):
+    if error_message is not None:
+        return gr.outputs.Textbox(text=error_message, placeholder="Error")
+    else:
+        return None
+
 iface = gr.Interface(
     fn=generate_output,
     inputs=[
         gr.inputs.Textbox(lines=1, label="Nombre", placeholder="Ingresa tu nombre"),
         gr.inputs.Textbox(lines=1, label="Fecha de Nacimiento", placeholder="DD/MM/AAAA"),
-        gr.inputs.Radio(choices, label="Selecciona una imagen:", thumbnails=[get_thumbnail(image, 50) for image in choices])
+        gr.inputs.Radio(choices, label="Selecciona una vidente:", thumbnails=[get_thumbnail(image, 50) for image in choices])
     ],
     outputs=[
-        gr.outputs.Video(label="Respuesta de Andrea (un minuto aproximadamente)").style(width=256),
+        gr.outputs.Video(label="Respuesta del Oraculo (un minuto aproximadamente)").style(width=256),
         gr.outputs.Textbox(label="Mensaje de error", type="text")
     ],
     title="Oraculo de Inteligencia Artifical v2.1",