Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -2,21 +2,45 @@ import gradio as gr
|
|
| 2 |
from kokoro import KPipeline
|
| 3 |
import numpy as np
|
| 4 |
import torch
|
|
|
|
| 5 |
|
| 6 |
-
# Detectar hardware
|
| 7 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 8 |
|
| 9 |
-
#
|
| 10 |
-
# Esto descargará automáticamente el modelo de 82M parámetros
|
| 11 |
pipeline = KPipeline(lang_code='e', device=device)
|
| 12 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 13 |
def tts_pro(text, voice_name):
|
| 14 |
if not text or not text.strip():
|
| 15 |
return None
|
| 16 |
|
| 17 |
try:
|
| 18 |
-
#
|
| 19 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 20 |
|
| 21 |
audio_segments = []
|
| 22 |
for _, _, audio in generator:
|
|
@@ -25,30 +49,80 @@ def tts_pro(text, voice_name):
|
|
| 25 |
if not audio_segments:
|
| 26 |
return None
|
| 27 |
|
| 28 |
-
#
|
| 29 |
final_audio = np.concatenate(audio_segments)
|
| 30 |
|
| 31 |
-
# Devolvemos el sample rate (24000) y el audio para Gradio
|
| 32 |
return (24000, final_audio)
|
| 33 |
|
| 34 |
except Exception as e:
|
| 35 |
print(f"Error en la generación: {e}")
|
| 36 |
return None
|
| 37 |
|
| 38 |
-
#
|
| 39 |
-
|
| 40 |
-
|
| 41 |
-
|
| 42 |
-
|
| 43 |
-
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
| 50 |
-
|
| 51 |
-
)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 52 |
|
| 53 |
if __name__ == "__main__":
|
| 54 |
-
demo.launch(
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 2 |
from kokoro import KPipeline
|
| 3 |
import numpy as np
|
| 4 |
import torch
|
| 5 |
+
import os
|
| 6 |
|
| 7 |
+
# Detectar hardware
|
| 8 |
device = "cuda" if torch.cuda.is_available() else "cpu"
|
| 9 |
|
| 10 |
+
# Inicializar pipeline - 'e' para español
|
|
|
|
| 11 |
pipeline = KPipeline(lang_code='e', device=device)
|
| 12 |
|
| 13 |
+
# Lista COMPLETA de voces del Space original
|
| 14 |
+
VOICES = [
|
| 15 |
+
"af_alloy.pt", "af_aoede.pt", "af_bella.pt", "af_heart.pt", "af_jessica.pt",
|
| 16 |
+
"af_kore.pt", "af_nicole.pt", "af_nova.pt", "af_river.pt", "af_sarah.pt",
|
| 17 |
+
"af_sky.pt", "am_adam.pt", "am_echo.pt", "am_eric.pt", "am_fenrir.pt",
|
| 18 |
+
"am_liam.pt", "am_michael.pt", "am_onyx.pt", "am_puck.pt", "bf_alice.pt",
|
| 19 |
+
"bf_isabella.pt", "bf_lily.pt", "bm_daniel.pt", "bm_fable.pt", "bm_george.pt",
|
| 20 |
+
"bm_lewis.pt", "ef_dora.pt", "em_alex.pt", "em_santa.pt", "ff_siwis.pt",
|
| 21 |
+
"hf_alpha.pt", "hf_beta.pt", "hm_omega.pt", "hm_psi.pt", "if_sara.pt",
|
| 22 |
+
"im_nicola.pt", "jf_alpha.pt", "jf_gongitsune.pt", "jf_nezumi.pt",
|
| 23 |
+
"jf_tebukuro.pt", "jm_kumo.pt", "pf_dora.pt", "pm_alex.pt", "pm_santa.pt",
|
| 24 |
+
"zf_xiaobei.pt", "zf_xiaoni.pt", "zf_xiaoyi.pt", "zm_yunjian.pt",
|
| 25 |
+
"zm_yunxi.pt", "zm_yunxia.pt", "zm_yunyang.pt"
|
| 26 |
+
]
|
| 27 |
+
|
| 28 |
def tts_pro(text, voice_name):
|
| 29 |
if not text or not text.strip():
|
| 30 |
return None
|
| 31 |
|
| 32 |
try:
|
| 33 |
+
# Verificar si el archivo de voz existe
|
| 34 |
+
voice_path = f"voices/{voice_name}"
|
| 35 |
+
|
| 36 |
+
# Si no está en voices/, buscar en el directorio actual
|
| 37 |
+
if not os.path.exists(voice_path):
|
| 38 |
+
voice_path = voice_name
|
| 39 |
+
|
| 40 |
+
print(f"Usando voz: {voice_path}")
|
| 41 |
+
|
| 42 |
+
# Generar audio - pasar la ruta del archivo .pt como voz
|
| 43 |
+
generator = pipeline(text, voice=voice_path, speed=1.0)
|
| 44 |
|
| 45 |
audio_segments = []
|
| 46 |
for _, _, audio in generator:
|
|
|
|
| 49 |
if not audio_segments:
|
| 50 |
return None
|
| 51 |
|
| 52 |
+
# Concatenar audio
|
| 53 |
final_audio = np.concatenate(audio_segments)
|
| 54 |
|
|
|
|
| 55 |
return (24000, final_audio)
|
| 56 |
|
| 57 |
except Exception as e:
|
| 58 |
print(f"Error en la generación: {e}")
|
| 59 |
return None
|
| 60 |
|
| 61 |
+
# Crear la interfaz
|
| 62 |
+
with gr.Blocks(title="Kokoro TTS - Aliah Plus", theme=gr.themes.Soft()) as demo:
|
| 63 |
+
gr.Markdown("# 🎤 Kokoro TTS v1.0")
|
| 64 |
+
gr.Markdown("### Texto a Voz con múltiples voces")
|
| 65 |
+
gr.Markdown("Escribe el texto y selecciona una voz para generar audio.")
|
| 66 |
+
|
| 67 |
+
with gr.Row():
|
| 68 |
+
with gr.Column():
|
| 69 |
+
texto_input = gr.Textbox(
|
| 70 |
+
label="Texto a convertir",
|
| 71 |
+
placeholder="Escribe aquí el texto en español...",
|
| 72 |
+
lines=4,
|
| 73 |
+
max_lines=10
|
| 74 |
+
)
|
| 75 |
+
|
| 76 |
+
voz_seleccionada = gr.Dropdown(
|
| 77 |
+
choices=VOICES,
|
| 78 |
+
label="Selecciona una voz",
|
| 79 |
+
value="af_bella.pt",
|
| 80 |
+
info="Elige entre más de 50 voces diferentes"
|
| 81 |
+
)
|
| 82 |
+
|
| 83 |
+
generar_btn = gr.Button("🎵 Generar Audio", variant="primary")
|
| 84 |
+
|
| 85 |
+
with gr.Column():
|
| 86 |
+
audio_output = gr.Audio(
|
| 87 |
+
label="Audio Generado",
|
| 88 |
+
type="numpy",
|
| 89 |
+
interactive=False
|
| 90 |
+
)
|
| 91 |
+
|
| 92 |
+
# Ejemplos
|
| 93 |
+
gr.Markdown("### Ejemplos rápidos:")
|
| 94 |
+
ejemplos = gr.Examples(
|
| 95 |
+
examples=[
|
| 96 |
+
["Hola, ¿cómo estás? Soy una voz generada por inteligencia artificial.", "af_bella.pt"],
|
| 97 |
+
["Bienvenido al sistema de texto a voz más avanzado.", "af_nova.pt"],
|
| 98 |
+
["La tecnología de síntesis de voz ha avanzado mucho.", "am_adam.pt"]
|
| 99 |
+
],
|
| 100 |
+
inputs=[texto_input, voz_seleccionada],
|
| 101 |
+
outputs=audio_output,
|
| 102 |
+
fn=tts_pro,
|
| 103 |
+
cache_examples=False
|
| 104 |
+
)
|
| 105 |
+
|
| 106 |
+
# Conectar el botón
|
| 107 |
+
generar_btn.click(
|
| 108 |
+
fn=tts_pro,
|
| 109 |
+
inputs=[texto_input, voz_seleccionada],
|
| 110 |
+
outputs=audio_output
|
| 111 |
+
)
|
| 112 |
+
|
| 113 |
+
# Información adicional
|
| 114 |
+
gr.Markdown("---")
|
| 115 |
+
gr.Markdown("""
|
| 116 |
+
### 📝 Notas:
|
| 117 |
+
- Las voces están en formato `.pt` y se cargan desde la carpeta `voices/`
|
| 118 |
+
- Tiempo de generación: ~5-10 segundos dependiendo del texto
|
| 119 |
+
- Formato de audio: 24kHz, mono
|
| 120 |
+
- Modelo: Kokoro TTS v1.0
|
| 121 |
+
""")
|
| 122 |
|
| 123 |
if __name__ == "__main__":
|
| 124 |
+
demo.launch(
|
| 125 |
+
share=False,
|
| 126 |
+
server_name="0.0.0.0" if os.getenv('SPACE_ID') else None,
|
| 127 |
+
server_port=7860
|
| 128 |
+
)
|