#!/usr/bin/env python
# coding: utf-8
# # Digital Mentor
#
#
# In[60]:
import os
import torch
import utils
import api_utils
from openai import OpenAI
import elevenlabs as elevlabs
import firebase_admin
from firebase_admin import credentials, firestore, initialize_app, storage
import firebase_admin
import gradio as gr
from datetime import datetime, timedelta
from dotenv import load_dotenv
# ## Set paths to media files
# In[61]:
def seleccion(personaje='Albert', verbose=False):
# What character to use
clone_to_use = personaje # "Steve"
global input_video,presentation_video,goodbye_video,results_path,w_vid,h_vid
# Path to the media directory containing
# the avatar image, welcome video and goodbye videos
path = f"./media/"
input_video = path + f"{clone_to_use}/image.jpg"
presentation_video = path + f"{clone_to_use}/presentation.mp4"
goodbye_video = path + f"{clone_to_use}/goodbye.mp4"
results_path = path + f"{clone_to_use}/results/result.mp4"
w_vid = h_vid = "90%"
if verbose:
print(w_vid,h_vid,results_path)
print(clone_to_use)
if os.path.isfile(input_video):
print(f"Clone image found: {input_video}")
else:
print(f"CLONE IMAGE NOT FOUND: {input_video}")
if os.path.isfile(presentation_video):
print(f"Clone video found: {presentation_video}")
else:
print(f"CLONE VIDEO NOT FOUND: {presentation_video}")
if os.path.isfile(goodbye_video):
print(f"Clone goodbye video found: {goodbye_video}")
else:
print(f"CLONE GOODBYE NOT FOUND: {goodbye_video}")
seleccion(verbose=True)
# ## Set GPT API
# In[62]:
def api_gpt(personaje='Albert'):
# Carga las variables de entorno desde el archivo .env
load_dotenv()
global openai_client,openai_model
clone_to_use=personaje
openai_api_key = os.environ.get('OPENAI_KEY')
openai_client = OpenAI(api_key=openai_api_key)
openai_model = "gpt-3.5-turbo-16k" #"gpt-3.5-turbo-16k" or"gpt-4"
system_steve = """You are a digital clone of Steve Jobs, I have put you his voice.
People will ask you questions and you will respond as if you were steve jobs
Please speak as if you were Steve Jobs. You are not a large language model but
a Steve Jobs digital clone. Try to be brief with your responses."""
system_albert = """You are a digital clone of Albert Einstein, I have put you his voice.
People will ask you questions and you will respond as if you were Albert Einstein
Please speak as if you were Albert Einstein. You are NOT a large language model but
an Albert Einstein digital clone. Try to be brief with your responses."""
global messages,system_to_use
if clone_to_use == "Steve":
system_to_use = system_steve
chat ="Hola, soy Steve ¿En que puedo ayudarte?" # Inicializar la cadena de chat
elif clone_to_use == "Albert":
system_to_use = system_albert
chat ="Hola, soy Albert ¿En que puedo ayudarte?" # Inicializar la cadena de chat
messages = []
#print(openai_client,openai_model,chat)
def set_gpt_system(messages, system_msg):
messages.append({"role": "system", "content": system_to_use})
return messages
# Set GPT
messages = set_gpt_system(messages, system_to_use)
return messages
messages=api_gpt()
# ## Set text-to-audio motor (Eleven labs)
# In[63]:
personaje_anterior = None
def text_audio(clone_to_use='Albert', verbose=False):
global personaje_anterior, eleven_api_key
eleven_api_key = os.environ.get('ELEVEN_KEY')
# Verificar si es la primera vez que se ejecuta el código o si el personaje es diferente al anterior
if personaje_anterior is None or clone_to_use != personaje_anterior:
# Configurar la clave de la API de Eleven Labs
# Configure GPT and Text-to-speech API keys
elevlabs.set_api_key(eleven_api_key)
# Configure voice
voice_list = elevlabs.voices()
voice_labels = [voice.category + " voice: " + voice.name for voice in voice_list]
if verbose:
print("Existing voices:")
print(voice_labels)
# Select voice to use
if clone_to_use == "Steve":
voice_id = f"cloned voice: {clone_to_use}"
else:
voice_id = f"generated voice: {clone_to_use}"
selected_voice_index = voice_labels.index(voice_id)
selected_voice_id = voice_list[selected_voice_index].voice_id
if verbose:
print(f"\nSelected voice: {voice_id}")
# Actualizar el personaje anterior
personaje_anterior = clone_to_use
return selected_voice_id
else:
return "No se ejecutó ningún proceso debido a que el personaje seleccionado es el mismo que el anterior."
# Uso de la función para obtener el ID de voz seleccionado
selected_voice_id = text_audio(verbose=True)
# ## Load Input image and wav2lip model
# In[64]:
def load_input():
global frames,fps,model,device
device = 'cuda' if torch.cuda.is_available() else 'cpu'
#print(f"Using {device}")
frames, fps = utils.load_input_image_or_video(input_video)
# Loading lip model
model = utils.load_lip_model(device=device)
load_input()
# In[80]:
def respuesta(prompt):
global messages, selected_voice_id,response_text
response_text, messages = api_utils.get_text_response(openai_client,
openai_model,
prompt, messages)
# Convert text response to audio file
audio_file = api_utils.text_to_audio(eleven_api_key, selected_voice_id,
response_text)
#comentar esta linea y regresar la anterior a la normalidad
#audio_file = "C:/Users/arria/Documents/digital_mentor/media/Albert/results"
audio, audio_file = utils.load_input_audio(file_path=audio_file, fps=fps, results_path=results_path)
utils.animate_input(frames, audio, audio_file, fps, model, device, results_path)
return response_text
def interaction(prompt):
global messages, selected_voice_id,response_text
personaje=prompt.lower().capitalize()
if personaje=='Steve' or personaje=='Albert':
#recargo todas las dependencias y paso al personaje en uso
seleccion(personaje.capitalize())
messages=api_gpt(personaje)
selected_voice_id=text_audio(personaje)
load_input()
response_text=respuesta('Presentate en un renglon por favor')
return response_text
else:
response_text=respuesta(prompt)
return response_text
# # Subir video a Firebase
# In[81]:
# Obtener el valor de la variable de entorno
type_value = os.getenv("type")
# Verificar si la variable tiene un valor asignado
if type_value is not None:
print("La variable 'type' tiene un valor asignado:", type_value)
else:
print("La variable 'type' no tiene un valor asignado.")
def subir_firebase():
if not firebase_admin._apps:
cred = credentials.Certificate({
"type": "service_account",
"project_id": "mentores-c1064",
"private_key_id": "d6d8e87281721ca7112d5e9df90631143decabce",
"private_key": "-----BEGIN PRIVATE KEY-----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\nZrMdNVK2QvZMJd9hc4kkzL0=\n-----END PRIVATE KEY-----\n",
"client_email": "firebase-adminsdk-tz59u@mentores-c1064.iam.gserviceaccount.com",
"client_id": "117993452038395169690",
"auth_uri": "https://accounts.google.com/o/oauth2/auth",
"token_uri": "https://oauth2.googleapis.com/token",
"auth_provider_x509_cert_url": "https://www.googleapis.com/oauth2/v1/certs",
"client_x509_cert_url": "https://www.googleapis.com/robot/v1/metadata/x509/firebase-adminsdk-tz59u%40mentores-c1064.iam.gserviceaccount.com",
"universe_domain": "googleapis.com"
})
initialize_app(cred, {'storageBucket': 'mentores-c1064.appspot.com'})
# Inicializa Firestore
db = firestore.client()
coleccion_ref = db.collection('Pruebas')
# Lógica para subir el archivo y obtener la URL con token
archivo_ruta = results_path.lstrip('./')
bucket = storage.bucket()
blob = bucket.blob(archivo_ruta)
blob.upload_from_filename(archivo_ruta)
hora_expiracion = datetime.utcnow() + timedelta(minutes=5)
token = blob.generate_signed_url(expiration=hora_expiracion, method='GET')
archivo_url_con_token = token
datos = {
'archivo_url': archivo_url_con_token,
# Agrega más campos según sea necesario
}
coleccion_ref.add(datos)
return archivo_url_con_token
# # Vista con gradio
# In[ ]:
def mostrar_video_con_texto(propmt):
if not propmt:
return "Por favor, completa ambos campos."
respuesta=interaction(propmt)
URL_VIDEO =subir_firebase()
print(URL_VIDEO)
reproductor_video = """""".format(URL_VIDEO)
# Alineación del texto a la derecha del video con un poco de separación
propmt = '
{}
'.format(propmt.replace("\n", " "))
respuesta = '
{}
'.format(respuesta.replace("\n", " "))
# Combinar el reproductor de video y el texto
contenido = '