Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,68 +1,48 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
import os
|
| 3 |
import asyncio
|
| 4 |
-
import
|
| 5 |
-
from
|
| 6 |
-
from pydub import AudioSegment
|
| 7 |
-
import io
|
| 8 |
|
| 9 |
-
|
| 10 |
-
pipe = pipeline(
|
| 11 |
-
"text-generation",
|
| 12 |
-
model="OpenAssistant/llama2-7b-orca-8k-3319",
|
| 13 |
-
device="cpu",
|
| 14 |
-
torch_dtype=torch.float32
|
| 15 |
-
)
|
| 16 |
|
| 17 |
-
|
| 18 |
-
|
| 19 |
-
# Mock: Guarda un audio vacío (implementa TTS real aquí)
|
| 20 |
-
audio_path = "output.mp3"
|
| 21 |
-
AudioSegment.silent(duration=1000).export(audio_path, format="mp3")
|
| 22 |
-
return audio_path
|
| 23 |
|
| 24 |
-
async def synthesize(
|
| 25 |
-
#
|
| 26 |
-
|
| 27 |
-
|
| 28 |
-
|
| 29 |
-
return "Error: No hay texto para procesar", None
|
| 30 |
-
|
| 31 |
-
# Genera diálogo en el idioma del texto
|
| 32 |
-
prompt = f"Convierte esto en un diálogo de podcast (en {language}): {text}"
|
| 33 |
-
output = pipe(prompt, max_length=1000)
|
| 34 |
-
conversation = output[0]["generated_text"]
|
| 35 |
-
|
| 36 |
-
# Convierte a audio (simulado)
|
| 37 |
-
audio_path = text_to_speech(conversation, language)
|
| 38 |
-
return conversation, audio_path
|
| 39 |
|
| 40 |
-
|
| 41 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 42 |
|
| 43 |
-
# Interfaz
|
| 44 |
with gr.Blocks(theme='gstaff/sketch') as demo:
|
| 45 |
-
gr.Markdown("#
|
| 46 |
-
gr.Markdown("### Soporta español/otros idiomas (sin GPU/API)")
|
| 47 |
-
|
| 48 |
with gr.Group():
|
| 49 |
text_url = gr.Textbox(label="URL del artículo (opcional)", placeholder="Ej: https://example.com")
|
| 50 |
-
text_input = gr.Textbox(label="O pega el texto aquí", lines=5, placeholder="
|
| 51 |
-
|
| 52 |
-
label="Idioma",
|
| 53 |
-
choices=["es", "en", "fr", "de"],
|
| 54 |
-
value="es"
|
| 55 |
-
)
|
| 56 |
-
btn = gr.Button("Generar Podcast", variant="primary")
|
| 57 |
|
| 58 |
with gr.Row():
|
| 59 |
-
conv_display = gr.Textbox(label="Conversación
|
| 60 |
-
aud = gr.Audio(label="Podcast", interactive=False)
|
| 61 |
|
| 62 |
btn.click(
|
| 63 |
synthesize_sync,
|
| 64 |
-
inputs=[text_url, text_input
|
| 65 |
outputs=[conv_display, aud]
|
| 66 |
)
|
| 67 |
|
| 68 |
-
demo.launch()
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
import os
|
| 3 |
import asyncio
|
| 4 |
+
from conver import ConversationConfig, URLToAudioConverter
|
| 5 |
+
from dotenv import load_dotenv
|
|
|
|
|
|
|
| 6 |
|
| 7 |
+
load_dotenv()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 8 |
|
| 9 |
+
def synthesize_sync(article_url, text_input):
|
| 10 |
+
return asyncio.run(synthesize(article_url, text_input))
|
|
|
|
|
|
|
|
|
|
|
|
|
| 11 |
|
| 12 |
+
async def synthesize(article_url, text_input):
|
| 13 |
+
# Si hay texto manual, úsalo; si no, usa la URL
|
| 14 |
+
input_text = text_input if text_input else article_url
|
| 15 |
+
if not input_text:
|
| 16 |
+
return "Ingresa una URL o texto", None
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 17 |
|
| 18 |
+
try:
|
| 19 |
+
config = ConversationConfig()
|
| 20 |
+
converter = URLToAudioConverter(config, llm_api_key=os.environ.get("TOGETHER_API_KEY"))
|
| 21 |
+
|
| 22 |
+
output_file, conversation = await converter.url_to_audio(
|
| 23 |
+
input_text,
|
| 24 |
+
"es-ES-AlvaroNeural", # Voz en español
|
| 25 |
+
"es-ES-ElviraNeural" # Segunda voz en español
|
| 26 |
+
)
|
| 27 |
+
return conversation, output_file
|
| 28 |
+
except Exception as e:
|
| 29 |
+
return f"Error: {str(e)}", None
|
| 30 |
|
|
|
|
| 31 |
with gr.Blocks(theme='gstaff/sketch') as demo:
|
| 32 |
+
gr.Markdown("# Convertir Artículo/Texto en Podcast")
|
|
|
|
|
|
|
| 33 |
with gr.Group():
|
| 34 |
text_url = gr.Textbox(label="URL del artículo (opcional)", placeholder="Ej: https://example.com")
|
| 35 |
+
text_input = gr.Textbox(label="O pega el texto aquí", lines=5, placeholder="Texto en español...")
|
| 36 |
+
btn = gr.Button("Podcastify", variant="primary")
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 37 |
|
| 38 |
with gr.Row():
|
| 39 |
+
conv_display = gr.Textbox(label="Conversación", interactive=False, lines=10)
|
| 40 |
+
aud = gr.Audio(label="Podcast generado", interactive=False)
|
| 41 |
|
| 42 |
btn.click(
|
| 43 |
synthesize_sync,
|
| 44 |
+
inputs=[text_url, text_input],
|
| 45 |
outputs=[conv_display, aud]
|
| 46 |
)
|
| 47 |
|
| 48 |
+
demo.queue().launch()
|