Spaces:
Build error
Build error
Create app.py
Browse files
app.py
ADDED
|
@@ -0,0 +1,60 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
import gradio as gr
|
| 2 |
+
import whisper
|
| 3 |
+
from transformers import MBartForConditionalGeneration, MBart50TokenizerFast
|
| 4 |
+
|
| 5 |
+
## Cargar modelos de Whisper
|
| 6 |
+
whisper_model = whisper.load_model("base.en")
|
| 7 |
+
|
| 8 |
+
## Cargar modelos de MBart
|
| 9 |
+
translation_model = MBartForConditionalGeneration.from_pretrained("SnypzZz/Llama2-13b-Language-translate")
|
| 10 |
+
tokenizer = MBart50TokenizerFast.from_pretrained("SnypzZz/Llama2-13b-Language-translate", src_lang="en_XX")
|
| 11 |
+
|
| 12 |
+
## Función para transcribir y traducir el audio
|
| 13 |
+
def transcribe_translate(audio_file, target_language):
|
| 14 |
+
# Transcribir audio con Whisper (aquí se usa la variable whisper_model)
|
| 15 |
+
transcription = whisper_model.transcribe(audio_file, language="english")["text"]
|
| 16 |
+
|
| 17 |
+
# Traducir texto a idioma seleccionado (aquí se usa translation_model y tokenizer)
|
| 18 |
+
model_inputs = tokenizer(transcription, return_tensors="pt")
|
| 19 |
+
generated_tokens = translation_model.generate(
|
| 20 |
+
**model_inputs,
|
| 21 |
+
forced_bos_token_id=tokenizer.lang_code_to_id[target_language]
|
| 22 |
+
)
|
| 23 |
+
translated_text = tokenizer.batch_decode(generated_tokens, skip_special_tokens=True)[0]
|
| 24 |
+
return translated_text.strip("[]' ")
|
| 25 |
+
|
| 26 |
+
## Interfaz de Gradio
|
| 27 |
+
# Está creado en filas para "organizar" la distribución de cada caja
|
| 28 |
+
with gr.Blocks(theme="Nymbo/Nymbo_Theme") as app:
|
| 29 |
+
# Título
|
| 30 |
+
gr.Markdown("## Transcripción y Traducción de Audio")
|
| 31 |
+
|
| 32 |
+
# Primera fila -> Input de audio y elección del idioma
|
| 33 |
+
with gr.Row():
|
| 34 |
+
# Audio
|
| 35 |
+
audio_input = gr.Audio(label="Subir o grabar audio en `inglés` exclusivamente`", sources=["upload", "microphone"], type="filepath")
|
| 36 |
+
|
| 37 |
+
# Elección de idioma
|
| 38 |
+
language_dropdown = gr.Dropdown(
|
| 39 |
+
["de_DE", "es_XX", "fr_XX", "sv_SE", "ru_RU"],
|
| 40 |
+
label="Selecciona el idioma de traducción",
|
| 41 |
+
value="es_XX"
|
| 42 |
+
)
|
| 43 |
+
|
| 44 |
+
# Segunda fila -> Botón y salida de texto traducido
|
| 45 |
+
with gr.Row():
|
| 46 |
+
# Boton
|
| 47 |
+
translate_button = gr.Button("Transcribir y Traducir")
|
| 48 |
+
|
| 49 |
+
# Caja de texto (output)
|
| 50 |
+
translation_output = gr.Textbox(label="Texto Traducido")
|
| 51 |
+
|
| 52 |
+
# Configuración botón
|
| 53 |
+
translate_button.click(
|
| 54 |
+
transcribe_translate,
|
| 55 |
+
inputs=[audio_input, language_dropdown],
|
| 56 |
+
outputs=translation_output
|
| 57 |
+
)
|
| 58 |
+
|
| 59 |
+
##Iniciar aplicacion
|
| 60 |
+
app.queue().launch()
|