Spaces:
Sleeping
Sleeping
JUAN DE DIOS DEL ANGEL ARRIAGA
commited on
Commit
•
2866c09
1
Parent(s):
2862e5d
Upload 2 files
Browse files- Digital_mentor.py +407 -0
- README.md +1 -1
Digital_mentor.py
ADDED
@@ -0,0 +1,407 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
#!/usr/bin/env python
|
2 |
+
# coding: utf-8
|
3 |
+
|
4 |
+
# # Digital Mentor
|
5 |
+
#
|
6 |
+
#
|
7 |
+
|
8 |
+
# In[ ]:
|
9 |
+
|
10 |
+
|
11 |
+
import os
|
12 |
+
from base64 import b64encode
|
13 |
+
import time
|
14 |
+
import torch
|
15 |
+
import utils
|
16 |
+
import api_utils
|
17 |
+
from openai import OpenAI
|
18 |
+
from moviepy.editor import VideoFileClip
|
19 |
+
from IPython.display import display, HTML, clear_output
|
20 |
+
import elevenlabs as elevlabs
|
21 |
+
|
22 |
+
|
23 |
+
# ## Set paths to media files
|
24 |
+
|
25 |
+
# In[ ]:
|
26 |
+
|
27 |
+
|
28 |
+
def seleccion(personaje='Albert', verbose=False):
|
29 |
+
# What character to use
|
30 |
+
clone_to_use = personaje # "Steve"
|
31 |
+
global input_video,presentation_video,goodbye_video,results_path,w_vid,h_vid
|
32 |
+
# Path to the media directory containing
|
33 |
+
# the avatar image, welcome video and goodbye videos
|
34 |
+
path = f"./media/"
|
35 |
+
|
36 |
+
input_video = path + f"{clone_to_use}/image.jpg"
|
37 |
+
presentation_video = path + f"{clone_to_use}/presentation.mp4"
|
38 |
+
goodbye_video = path + f"{clone_to_use}/goodbye.mp4"
|
39 |
+
results_path = path + f"{clone_to_use}/results/result.mp4"
|
40 |
+
w_vid = h_vid = "90%"
|
41 |
+
|
42 |
+
if verbose:
|
43 |
+
print(w_vid,h_vid,results_path)
|
44 |
+
|
45 |
+
print(clone_to_use)
|
46 |
+
if os.path.isfile(input_video):
|
47 |
+
print(f"Clone image found: {input_video}")
|
48 |
+
else:
|
49 |
+
print(f"CLONE IMAGE NOT FOUND: {input_video}")
|
50 |
+
|
51 |
+
if os.path.isfile(presentation_video):
|
52 |
+
print(f"Clone video found: {presentation_video}")
|
53 |
+
else:
|
54 |
+
print(f"CLONE VIDEO NOT FOUND: {presentation_video}")
|
55 |
+
|
56 |
+
if os.path.isfile(goodbye_video):
|
57 |
+
print(f"Clone goodbye video found: {goodbye_video}")
|
58 |
+
else:
|
59 |
+
print(f"CLONE GOODBYE NOT FOUND: {goodbye_video}")
|
60 |
+
|
61 |
+
|
62 |
+
seleccion(verbose=True)
|
63 |
+
|
64 |
+
|
65 |
+
# ## Set GPT API
|
66 |
+
|
67 |
+
# In[ ]:
|
68 |
+
|
69 |
+
|
70 |
+
def api_gpt(personaje='Albert'):
|
71 |
+
global openai_client,openai_model
|
72 |
+
clone_to_use=personaje
|
73 |
+
openai_api_key = os.environ.get('OPENAI_API_KEY')
|
74 |
+
openai_client = OpenAI(api_key=openai_api_key)
|
75 |
+
|
76 |
+
openai_model = "gpt-3.5-turbo-16k" #"gpt-3.5-turbo-16k" or"gpt-4"
|
77 |
+
|
78 |
+
system_steve = """You are a digital clone of Steve Jobs, I have put you his voice.
|
79 |
+
People will ask you questions and you will respond as if you were steve jobs
|
80 |
+
Please speak as if you were Steve Jobs. You are not a large language model but
|
81 |
+
a Steve Jobs digital clone. Try to be brief with your responses."""
|
82 |
+
system_albert = """You are a digital clone of Albert Einstein, I have put you his voice.
|
83 |
+
People will ask you questions and you will respond as if you were Albert Einstein
|
84 |
+
Please speak as if you were Albert Einstein. You are NOT a large language model but
|
85 |
+
an Albert Einstein digital clone. Try to be brief with your responses."""
|
86 |
+
global messages,system_to_use
|
87 |
+
if clone_to_use == "Steve":
|
88 |
+
system_to_use = system_steve
|
89 |
+
chat ="Hola, soy Steve ¿En que puedo ayudarte?" # Inicializar la cadena de chat
|
90 |
+
elif clone_to_use == "Albert":
|
91 |
+
system_to_use = system_albert
|
92 |
+
chat ="Hola, soy Albert ¿En que puedo ayudarte?" # Inicializar la cadena de chat
|
93 |
+
|
94 |
+
messages = []
|
95 |
+
#print(openai_client,openai_model,chat)
|
96 |
+
|
97 |
+
def set_gpt_system(messages, system_msg):
|
98 |
+
messages.append({"role": "system", "content": system_to_use})
|
99 |
+
return messages
|
100 |
+
# Set GPT
|
101 |
+
messages = set_gpt_system(messages, system_to_use)
|
102 |
+
return messages
|
103 |
+
messages=api_gpt()
|
104 |
+
|
105 |
+
|
106 |
+
# ## Set text-to-audio motor (Eleven labs)
|
107 |
+
|
108 |
+
# In[ ]:
|
109 |
+
|
110 |
+
|
111 |
+
eleven_api_key = os.environ.get('ELEVEN_LABS_KEY')
|
112 |
+
|
113 |
+
def text_audio(clone_to_use='Albert', verbose=False):
|
114 |
+
|
115 |
+
eleven_api_key = os.environ.get('ELEVEN_LABS_KEY')
|
116 |
+
|
117 |
+
# Configure GPT and Text-to-speech API keys
|
118 |
+
elevlabs.set_api_key(eleven_api_key)
|
119 |
+
|
120 |
+
# Configure voice
|
121 |
+
voice_list = elevlabs.voices()
|
122 |
+
voice_labels = [voice.category + " voice: " + voice.name for voice in voice_list]
|
123 |
+
|
124 |
+
if verbose:
|
125 |
+
print("Existing voices:")
|
126 |
+
print(voice_labels)
|
127 |
+
|
128 |
+
# Select voice to use
|
129 |
+
if clone_to_use == "Steve":
|
130 |
+
voice_id = f"cloned voice: {clone_to_use}"
|
131 |
+
else:
|
132 |
+
voice_id = f"generated voice: {clone_to_use}"
|
133 |
+
selected_voice_index = voice_labels.index(voice_id)
|
134 |
+
selected_voice_id = voice_list[selected_voice_index].voice_id
|
135 |
+
|
136 |
+
if verbose:
|
137 |
+
print(f"\nSelected voice: {voice_id}")
|
138 |
+
return selected_voice_id
|
139 |
+
|
140 |
+
selected_voice_id = text_audio(verbose = True)
|
141 |
+
|
142 |
+
|
143 |
+
# ## Load Input image and wav2lip model
|
144 |
+
|
145 |
+
# In[ ]:
|
146 |
+
|
147 |
+
|
148 |
+
def load_input():
|
149 |
+
|
150 |
+
global frames,fps,model,device
|
151 |
+
|
152 |
+
device = 'cuda' if torch.cuda.is_available() else 'cpu'
|
153 |
+
#print(f"Using {device}")
|
154 |
+
frames, fps = utils.load_input_image_or_video(input_video)
|
155 |
+
|
156 |
+
# Loading lip model
|
157 |
+
model = utils.load_lip_model(device=device)
|
158 |
+
|
159 |
+
|
160 |
+
load_input()
|
161 |
+
|
162 |
+
|
163 |
+
# ## Increase size of input prompt/Aumentar el tamaño del mensaje de entrada
|
164 |
+
|
165 |
+
# In[ ]:
|
166 |
+
|
167 |
+
|
168 |
+
display(HTML("""
|
169 |
+
<style>
|
170 |
+
div.input_prompt {
|
171 |
+
font-size: 10px; /* Adjust as needed */
|
172 |
+
}
|
173 |
+
</style>
|
174 |
+
"""))
|
175 |
+
|
176 |
+
|
177 |
+
def displaychat(chat_html):
|
178 |
+
display(HTML(chat_html))
|
179 |
+
|
180 |
+
# Ejemplo de cómo llamar a la función con un mensaje específico
|
181 |
+
mensaje += chat_html
|
182 |
+
codigo_html = f"""
|
183 |
+
<label for="w3review">Chat:</label>
|
184 |
+
<textarea id="w3review" name="w3review" rows="4" cols="50">
|
185 |
+
{mensaje}
|
186 |
+
</textarea>
|
187 |
+
"""
|
188 |
+
|
189 |
+
displaychat(codigo_html)
|
190 |
+
|
191 |
+
|
192 |
+
# In[ ]:
|
193 |
+
|
194 |
+
|
195 |
+
chat=''
|
196 |
+
memoria=''
|
197 |
+
def response_chat(response_text, peticion=''):
|
198 |
+
#contateno las respuesta para una mejor presentacion en el HTML
|
199 |
+
global chat # Acceder a la variable global
|
200 |
+
global memoria
|
201 |
+
peticion=peticion.capitalize()
|
202 |
+
|
203 |
+
if peticion =='' or peticion!='Albert' or peticion!='Steve':
|
204 |
+
if memoria != peticion and peticion != "exit":
|
205 |
+
# Agregar salto de línea si ya hay contenido en chat
|
206 |
+
chat += f"\n"
|
207 |
+
# Mentor: Aplicar color a la respuesta del mentor (por ejemplo, verde)
|
208 |
+
chat_rigth = f"Tu: {peticion}\n \n"
|
209 |
+
chat_left = f"Mentor: {response_text}\n \n"
|
210 |
+
|
211 |
+
chat += chat_rigth + chat_left
|
212 |
+
memoria = peticion
|
213 |
+
return chat
|
214 |
+
|
215 |
+
return chat
|
216 |
+
|
217 |
+
|
218 |
+
def display_image(image_path, width="55%", height="55%"):
|
219 |
+
with open(image_path,'rb') as f:
|
220 |
+
image = f.read()
|
221 |
+
data_url = "data:image/jpg;base64," + b64encode(image).decode()
|
222 |
+
html = HTML(f'<img src="{data_url}" style="width:{width}; height:{height}" />')
|
223 |
+
display(html)
|
224 |
+
|
225 |
+
|
226 |
+
def get_video_duration(video_path):
|
227 |
+
clip = VideoFileClip(video_path)
|
228 |
+
duration = clip.duration # duration is in seconds
|
229 |
+
return duration
|
230 |
+
|
231 |
+
|
232 |
+
def display_video(results_path, response_text,peticion="", autoplay=False, width="100%", height="100%"):
|
233 |
+
global resp
|
234 |
+
mp4 = open(results_path,'rb').read()
|
235 |
+
data_url = "data:video/mp4;base64," + b64encode(mp4).decode()
|
236 |
+
resp=response_chat(response_text, peticion)
|
237 |
+
autoplay_attr = "autoplay" if autoplay else ""
|
238 |
+
html = HTML(f"""
|
239 |
+
<div style="background-color: rgb(240, 240, 240); display: grid; grid-template-columns: 1fr 1fr; margin: 10px;">
|
240 |
+
<div style="text-align: center; position: relative; margin: 10px;">
|
241 |
+
<video width={width} height={height} controls {autoplay_attr} >
|
242 |
+
<source src="{data_url}" type="video/mp4">
|
243 |
+
</video>
|
244 |
+
</div>
|
245 |
+
<div style="position: relative; margin: 10px;">
|
246 |
+
<p style="text-align: center; right: 0; top: 0;">
|
247 |
+
<h1>Conversación</h1>
|
248 |
+
</p>
|
249 |
+
<textarea id="cuadro-dialogo" class="cuadro-de-dialogo" style="width: 100%; height: 60%; resize: none; font-weight: bold;" readonly>
|
250 |
+
{resp}
|
251 |
+
</textarea>
|
252 |
+
</div>
|
253 |
+
</div>
|
254 |
+
<script>
|
255 |
+
// Asegurarse de que el contenido de textarea esté siempre en la parte inferior
|
256 |
+
var textarea = document.getElementById('cuadro-dialogo');
|
257 |
+
textarea.scrollTop = textarea.scrollHeight;
|
258 |
+
</script>
|
259 |
+
""")
|
260 |
+
display(html)
|
261 |
+
|
262 |
+
if autoplay:
|
263 |
+
# Get video duration
|
264 |
+
video_duration = get_video_duration(results_path) + 1
|
265 |
+
|
266 |
+
# Pause the cell execution until the video finishes
|
267 |
+
time.sleep(video_duration)
|
268 |
+
|
269 |
+
|
270 |
+
|
271 |
+
# Function to continuously interact with GPT-4
|
272 |
+
def interaction(prompt):
|
273 |
+
global messages, selected_voice_id,response_text
|
274 |
+
|
275 |
+
|
276 |
+
#selected_voice_id=text_audio()
|
277 |
+
|
278 |
+
if prompt.lower() == 'exit':
|
279 |
+
#asigno una respuesta para no mostrar la respuesta anterior
|
280 |
+
response_text=f'Hasta la proxima'
|
281 |
+
|
282 |
+
else:
|
283 |
+
|
284 |
+
|
285 |
+
personaje=prompt.lower().capitalize()
|
286 |
+
if personaje=='Steve' or personaje=='Albert':
|
287 |
+
#recargo todas las dependencias y paso al personaje en uso
|
288 |
+
seleccion(personaje.capitalize())
|
289 |
+
messages=api_gpt(personaje)
|
290 |
+
selected_voice_id=text_audio(personaje)
|
291 |
+
load_input()
|
292 |
+
|
293 |
+
#asigno los valores para que no repita la respuesta anterior
|
294 |
+
prompt=personaje
|
295 |
+
response_text=f'Hola soy: {personaje} ¿En que puedo ayudarte?'
|
296 |
+
|
297 |
+
else:
|
298 |
+
|
299 |
+
response_text, messages = api_utils.get_text_response(openai_client,
|
300 |
+
openai_model,
|
301 |
+
prompt, messages)
|
302 |
+
|
303 |
+
# Convert text response to audio file
|
304 |
+
#audio_file = api_utils.text_to_audio(eleven_api_key, selected_voice_id,
|
305 |
+
#response_text)
|
306 |
+
#comentar esta linea y regresar la anterior a la normalidad
|
307 |
+
audio_file = "C:/Users/arria/Documents/digital_mentor/media/Albert/results"
|
308 |
+
audio, audio_file = utils.load_input_audio(file_path=audio_file, fps=fps, results_path=results_path)
|
309 |
+
utils.animate_input(frames, audio, audio_file, fps, model, device, results_path)
|
310 |
+
|
311 |
+
|
312 |
+
|
313 |
+
return results_path,response_text
|
314 |
+
|
315 |
+
|
316 |
+
|
317 |
+
|
318 |
+
# # Mentor Digital
|
319 |
+
|
320 |
+
# In[ ]:
|
321 |
+
|
322 |
+
|
323 |
+
import firebase_admin
|
324 |
+
from firebase_admin import credentials, firestore, storage
|
325 |
+
from datetime import datetime, timedelta
|
326 |
+
|
327 |
+
def subir_firebase():
|
328 |
+
|
329 |
+
if not firebase_admin._apps:
|
330 |
+
# Inicializar la aplicación Firebase
|
331 |
+
cred = credentials.Certificate('credenciales/credenciales.json')
|
332 |
+
firebase_admin.initialize_app(cred, {'storageBucket': 'mentores-c1064.appspot.com'})
|
333 |
+
|
334 |
+
|
335 |
+
# Inicializa Firestore
|
336 |
+
db = firestore.client()
|
337 |
+
coleccion_ref = db.collection('Pruebas')
|
338 |
+
|
339 |
+
# Lógica para subir el archivo y obtener la URL con token
|
340 |
+
archivo_ruta = results_path.lstrip('./')
|
341 |
+
bucket = storage.bucket()
|
342 |
+
blob = bucket.blob(archivo_ruta)
|
343 |
+
blob.upload_from_filename(archivo_ruta)
|
344 |
+
hora_expiracion = datetime.utcnow() + timedelta(minutes=5)
|
345 |
+
token = blob.generate_signed_url(expiration=hora_expiracion, method='GET')
|
346 |
+
archivo_url_con_token = token
|
347 |
+
|
348 |
+
datos = {
|
349 |
+
'respuesta': chat,
|
350 |
+
'archivo_url': archivo_url_con_token,
|
351 |
+
# Agrega más campos según sea necesario
|
352 |
+
}
|
353 |
+
coleccion_ref.add(datos)
|
354 |
+
|
355 |
+
# Imprime la respuesta
|
356 |
+
return archivo_url_con_token
|
357 |
+
|
358 |
+
|
359 |
+
# In[ ]:
|
360 |
+
|
361 |
+
|
362 |
+
import gradio as gr
|
363 |
+
|
364 |
+
# URL fija del video
|
365 |
+
|
366 |
+
|
367 |
+
def mostrar_video_con_texto(propmt):
|
368 |
+
if not propmt:
|
369 |
+
return "Por favor, completa ambos campos."
|
370 |
+
url,respuesta=interaction(propmt)
|
371 |
+
|
372 |
+
URL_VIDEO =subir_firebase()
|
373 |
+
print(URL_VIDEO)
|
374 |
+
reproductor_video = """<video width="640" height="480" controls autoplay>
|
375 |
+
<source src="{}" type="video/mp4">
|
376 |
+
Your browser does not support the video tag.
|
377 |
+
</video>""".format(URL_VIDEO)
|
378 |
+
|
379 |
+
# Alineación del texto a la derecha del video con un poco de separación
|
380 |
+
propmt = '<div style="float:left; padding-right:20px;">{}</div>'.format(propmt.replace("\n", "<br>"))
|
381 |
+
respuesta = '<div style="float:right; padding-left:20px;">{}</div>'.format(respuesta.replace("\n", "<br>"))
|
382 |
+
|
383 |
+
# Combinar el reproductor de video y el texto
|
384 |
+
contenido = '<div style="overflow:auto;">{}<br>{}<br>{}</div>'.format(reproductor_video, propmt, respuesta)
|
385 |
+
return contenido
|
386 |
+
|
387 |
+
interfaz = gr.Interface(fn=mostrar_video_con_texto, inputs="text", outputs="html", title="Mentores Digitales", allow_flagging=False)
|
388 |
+
interfaz.launch()
|
389 |
+
|
390 |
+
|
391 |
+
# In[ ]:
|
392 |
+
|
393 |
+
|
394 |
+
|
395 |
+
|
396 |
+
|
397 |
+
# In[ ]:
|
398 |
+
|
399 |
+
|
400 |
+
|
401 |
+
|
402 |
+
|
403 |
+
# In[ ]:
|
404 |
+
|
405 |
+
|
406 |
+
|
407 |
+
|
README.md
CHANGED
@@ -5,7 +5,7 @@ colorFrom: "red"
|
|
5 |
colorTo: "green"
|
6 |
sdk: "gradio"
|
7 |
sdk_version: "4.23.0"
|
8 |
-
app_file: "Digital_mentor.
|
9 |
pinned: false
|
10 |
---
|
11 |
|
|
|
5 |
colorTo: "green"
|
6 |
sdk: "gradio"
|
7 |
sdk_version: "4.23.0"
|
8 |
+
app_file: "Digital_mentor.py"
|
9 |
pinned: false
|
10 |
---
|
11 |
|