Update app.py
Browse files
app.py
CHANGED
|
@@ -1,136 +1,63 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
-
from
|
|
|
|
| 3 |
import torch
|
| 4 |
-
import
|
|
|
|
| 5 |
|
| 6 |
-
#
|
| 7 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 8 |
|
| 9 |
-
|
| 10 |
-
|
| 11 |
-
|
| 12 |
-
print("✅ Modelo cargado correctamente")
|
| 13 |
|
| 14 |
-
|
| 15 |
-
|
| 16 |
-
|
| 17 |
-
"""
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
if not text or not text.strip():
|
| 19 |
-
return None, "⚠️
|
| 20 |
|
| 21 |
-
|
| 22 |
-
|
| 23 |
-
return None, "⚠️ El texto es muy largo. Máximo 500 caracteres."
|
| 24 |
|
| 25 |
try:
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
|
| 30 |
-
text=text,
|
| 31 |
-
file_path=output_path,
|
| 32 |
-
speed=speed
|
| 33 |
)
|
| 34 |
|
| 35 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 36 |
|
| 37 |
except Exception as e:
|
| 38 |
return None, f"❌ Error: {str(e)}"
|
| 39 |
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
.gradio-container {
|
| 43 |
-
max-width: 800px !important;
|
| 44 |
-
margin: 0 auto;
|
| 45 |
-
}
|
| 46 |
-
.title {
|
| 47 |
-
text-align: center;
|
| 48 |
-
color: #2c3e50;
|
| 49 |
-
}
|
| 50 |
-
"""
|
| 51 |
-
|
| 52 |
-
# Crear la interfaz
|
| 53 |
-
with gr.Blocks(css=custom_css, title="🎙️ TTS Español Latino") as demo:
|
| 54 |
-
|
| 55 |
-
gr.Markdown("""
|
| 56 |
-
<div class="title">
|
| 57 |
-
<h1>🎙️ Lector de Texto a Voz</h1>
|
| 58 |
-
<h3>Español Latino - Rápido & Gratis</h3>
|
| 59 |
-
</div>
|
| 60 |
-
|
| 61 |
-
Escribe cualquier texto y escúchalo en voz alta en segundos.
|
| 62 |
-
""")
|
| 63 |
|
| 64 |
with gr.Row():
|
| 65 |
-
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
)
|
| 72 |
-
|
| 73 |
-
with gr.Row():
|
| 74 |
-
speed_slider = gr.Slider(
|
| 75 |
-
minimum=0.5,
|
| 76 |
-
maximum=1.5,
|
| 77 |
-
value=1.0,
|
| 78 |
-
step=0.1,
|
| 79 |
-
label="⚡ Velocidad de voz"
|
| 80 |
-
)
|
| 81 |
-
|
| 82 |
-
generate_btn = gr.Button(
|
| 83 |
-
"🔊 Generar Voz",
|
| 84 |
-
variant="primary",
|
| 85 |
-
size="lg"
|
| 86 |
-
)
|
| 87 |
-
|
| 88 |
-
status_text = gr.Textbox(
|
| 89 |
-
label="Estado",
|
| 90 |
-
interactive=False,
|
| 91 |
-
value="Listo para generar audio"
|
| 92 |
-
)
|
| 93 |
-
|
| 94 |
-
with gr.Column(scale=1):
|
| 95 |
-
audio_output = gr.Audio(
|
| 96 |
-
label="🔊 Tu Audio",
|
| 97 |
-
type="filepath",
|
| 98 |
-
autoplay=False
|
| 99 |
-
)
|
| 100 |
-
|
| 101 |
-
gr.Markdown("""
|
| 102 |
-
### 💡 Consejos:
|
| 103 |
-
- **Máximo**: 500 caracteres
|
| 104 |
-
- Usa puntuación para pausas naturales
|
| 105 |
-
- Ajusta la velocidad si es necesario
|
| 106 |
-
- Funciona mejor con oraciones completas
|
| 107 |
-
""")
|
| 108 |
-
|
| 109 |
-
# Ejemplos rápidos
|
| 110 |
-
gr.Examples(
|
| 111 |
-
examples=[
|
| 112 |
-
["Hola, ¿cómo estás? Bienvenido a este lector de texto a voz."],
|
| 113 |
-
["El sol brilla intensamente sobre las playas de México."],
|
| 114 |
-
["La tecnología nos permite crear herramientas increíbles cada día."],
|
| 115 |
-
["¿Podrías repetir eso, por favor? No lo escuché bien."]
|
| 116 |
-
],
|
| 117 |
-
inputs=text_input,
|
| 118 |
-
label="🎯 Probar con ejemplos"
|
| 119 |
-
)
|
| 120 |
|
| 121 |
-
|
| 122 |
-
|
| 123 |
-
|
| 124 |
-
inputs=[text_input, speed_slider],
|
| 125 |
-
outputs=[audio_output, status_text]
|
| 126 |
-
)
|
| 127 |
|
| 128 |
-
|
| 129 |
-
text_input.change(
|
| 130 |
-
fn=lambda: ("", "⌨️ Escribiendo..."),
|
| 131 |
-
outputs=[audio_output, status_text]
|
| 132 |
-
)
|
| 133 |
|
| 134 |
-
# Lanzar
|
| 135 |
if __name__ == "__main__":
|
| 136 |
demo.launch()
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
+
from transformers import pipeline
|
| 3 |
+
from datasets import load_dataset
|
| 4 |
import torch
|
| 5 |
+
import soundfile as sf
|
| 6 |
+
import numpy as np
|
| 7 |
|
| 8 |
+
# Cargar modelo de TTS de Microsoft (pequeño y rápido)
|
| 9 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 10 |
|
| 11 |
+
print("🔄 Cargando modelo TTS...")
|
| 12 |
+
# Usamos un modelo específico para español más ligero
|
| 13 |
+
synthesizer = pipeline("text-to-speech", "microsoft/speecht5_tts", device=device)
|
|
|
|
| 14 |
|
| 15 |
+
# Cargar embeddings de speaker para variedad
|
| 16 |
+
embeddings_dataset = load_dataset("Matthijs/cmu-arctic-xvectors", split="validation")
|
| 17 |
+
speaker_embeddings = {
|
| 18 |
+
"hombre_1": torch.tensor(embeddings_dataset[7306]["xvector"]).unsqueeze(0),
|
| 19 |
+
"mujer_1": torch.tensor(embeddings_dataset[7300]["xvector"]).unsqueeze(0),
|
| 20 |
+
}
|
| 21 |
+
|
| 22 |
+
def text_to_speech(text, speaker="hombre_1"):
|
| 23 |
if not text or not text.strip():
|
| 24 |
+
return None, "⚠️ Ingresa texto"
|
| 25 |
|
| 26 |
+
if len(text) > 200:
|
| 27 |
+
return None, "⚠️ Máximo 200 caracteres"
|
|
|
|
| 28 |
|
| 29 |
try:
|
| 30 |
+
# Generar
|
| 31 |
+
speech = synthesizer(
|
| 32 |
+
text,
|
| 33 |
+
forward_params={"speaker_embeddings": speaker_embeddings[speaker]}
|
|
|
|
|
|
|
|
|
|
| 34 |
)
|
| 35 |
|
| 36 |
+
# Guardar
|
| 37 |
+
output_path = "output.wav"
|
| 38 |
+
sf.write(output_path, speech["audio"], samplerate=speech["sampling_rate"])
|
| 39 |
+
|
| 40 |
+
return output_path, "✅ ¡Listo!"
|
| 41 |
|
| 42 |
except Exception as e:
|
| 43 |
return None, f"❌ Error: {str(e)}"
|
| 44 |
|
| 45 |
+
with gr.Blocks(title="🎙️ TTS Local Español") as demo:
|
| 46 |
+
gr.Markdown("# 🎙️ Texto a Voz - Modelo Local")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 47 |
|
| 48 |
with gr.Row():
|
| 49 |
+
text_input = gr.Textbox(label="Texto", lines=3)
|
| 50 |
+
speaker_select = gr.Dropdown(
|
| 51 |
+
choices=["hombre_1", "mujer_1"],
|
| 52 |
+
value="hombre_1",
|
| 53 |
+
label="Voz"
|
| 54 |
+
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 55 |
|
| 56 |
+
btn = gr.Button("Generar", variant="primary")
|
| 57 |
+
audio = gr.Audio(label="Audio")
|
| 58 |
+
status = gr.Textbox(label="Estado")
|
|
|
|
|
|
|
|
|
|
| 59 |
|
| 60 |
+
btn.click(fn=text_to_speech, inputs=[text_input, speaker_select], outputs=[audio, status])
|
|
|
|
|
|
|
|
|
|
|
|
|
| 61 |
|
|
|
|
| 62 |
if __name__ == "__main__":
|
| 63 |
demo.launch()
|