Create app.py

app.py

@@ -0,0 +1,41 @@
+import gradio as gr
+import torch
+from transformers import pipeline
+import numpy as np
+import time
+
+pipe_base = pipeline("automatic-speech-recognition", model="aitor-medrano/lara-base-pushed")
+pipe_small = pipeline("automatic-speech-recognition", model="aitor-medrano/whisper-small-lara")
+
+def greet(grabacion, modelo="base"):
+
+    inicio = time.time()
+
+    sr, y = grabacion
+    # Pasamos el array de muestras a tipo NumPy de 32 bits
+    y = y.astype(np.float32)
+    y /= np.max(np.abs(y))
+
+    if modelo is not None and modelo == "base":
+        pipe = pipe_base
+    else:
+        modelo = "small"
+        pipe = pipe_small
+
+    result = modelo + ":" + pipe({"sampling_rate": sr, "raw": y})["text"]
+    fin = time.time()
+
+    return result, fin - inicio
+
+demo = gr.Interface(fn=greet,
+        inputs=[
+                gr.Audio(),
+                gr.Dropdown(
+                    ["base", "small"], label="Modelo", info="Modelos de Lara entrenados"
+                )
+        ],
+        outputs=[
+            gr.Text(label="Salida"),
+            gr.Number(label="Tiempo")
+        ])
+demo.launch()