Spaces:
Sleeping
Sleeping
gggg
Browse files
app.py
CHANGED
|
@@ -10,24 +10,26 @@ from transformers import MarianMTModel, MarianTokenizer
|
|
| 10 |
|
| 11 |
|
| 12 |
# Cargar el modelo Whisper-small y bark
|
| 13 |
-
|
| 14 |
"""bark = pipeline("text-to-speech", model="suno/bark")"""
|
| 15 |
|
| 16 |
|
| 17 |
# Cargar el tokenizador y el modelo para espa帽ol a ingl茅s
|
| 18 |
-
|
| 19 |
-
tokenizer = MarianTokenizer.from_pretrained(model_name)
|
| 20 |
-
model = MarianMTModel.from_pretrained(model_name)
|
| 21 |
|
| 22 |
|
| 23 |
# Funci贸n para transcribir el audio y traducir el audio de entrada
|
| 24 |
def transcribir_audio(audio):
|
| 25 |
# Usamos el pipeline de Hugging Face para la transcripci贸n
|
| 26 |
-
|
|
|
|
| 27 |
return result["text"]
|
| 28 |
|
| 29 |
|
| 30 |
def traducir_texto(texto):
|
|
|
|
|
|
|
|
|
|
| 31 |
# Tokenizar el texto
|
| 32 |
inputs = tokenizer(texto, return_tensors="pt", padding=True, truncation=True)
|
| 33 |
# Generar la traducci贸n
|
|
@@ -46,9 +48,20 @@ def generar_audio(text):
|
|
| 46 |
temp_wav = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")
|
| 47 |
write(temp_wav.name, 24000, (audio_array * 32767).astype(np.int16))
|
| 48 |
return temp_wav.name
|
| 49 |
-
|
| 50 |
|
| 51 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
def process_audio(audio_file):
|
| 53 |
try:
|
| 54 |
# Paso 1: Transcripci贸n con Whisper
|
|
|
|
| 10 |
|
| 11 |
|
| 12 |
# Cargar el modelo Whisper-small y bark
|
| 13 |
+
|
| 14 |
"""bark = pipeline("text-to-speech", model="suno/bark")"""
|
| 15 |
|
| 16 |
|
| 17 |
# Cargar el tokenizador y el modelo para espa帽ol a ingl茅s
|
| 18 |
+
|
|
|
|
|
|
|
| 19 |
|
| 20 |
|
| 21 |
# Funci贸n para transcribir el audio y traducir el audio de entrada
|
| 22 |
def transcribir_audio(audio):
|
| 23 |
# Usamos el pipeline de Hugging Face para la transcripci贸n
|
| 24 |
+
transcribir = pipeline("automatic-speech-recognition", model="openai/whisper-small")
|
| 25 |
+
result = transcribir(audio)
|
| 26 |
return result["text"]
|
| 27 |
|
| 28 |
|
| 29 |
def traducir_texto(texto):
|
| 30 |
+
model_name = "Helsinki-NLP/opus-mt-es-en"
|
| 31 |
+
tokenizer = MarianTokenizer.from_pretrained(model_name)
|
| 32 |
+
model = MarianMTModel.from_pretrained(model_name)
|
| 33 |
# Tokenizar el texto
|
| 34 |
inputs = tokenizer(texto, return_tensors="pt", padding=True, truncation=True)
|
| 35 |
# Generar la traducci贸n
|
|
|
|
| 48 |
temp_wav = tempfile.NamedTemporaryFile(delete=False, suffix=".wav")
|
| 49 |
write(temp_wav.name, 24000, (audio_array * 32767).astype(np.int16))
|
| 50 |
return temp_wav.name
|
|
|
|
| 51 |
|
| 52 |
|
| 53 |
+
"""def process_audio(audio_file):
|
| 54 |
+
try:
|
| 55 |
+
# Paso 1: Transcripci贸n y traducci贸n con Whisper
|
| 56 |
+
transcripcion_traducida = transcribir_audio(audio_file)
|
| 57 |
+
|
| 58 |
+
# Paso 2: Generaci贸n de audio con Bark
|
| 59 |
+
audio_sintetizado = generar_audio(transcripcion_traducida)
|
| 60 |
+
|
| 61 |
+
return transcripcion_traducida, audio_sintetizado
|
| 62 |
+
except Exception as e:
|
| 63 |
+
return str(e), None"""
|
| 64 |
+
|
| 65 |
def process_audio(audio_file):
|
| 66 |
try:
|
| 67 |
# Paso 1: Transcripci贸n con Whisper
|