Create app.py

app.py

@@ -0,0 +1,16 @@
+import gradio as gr
+import numpy as np
+import torchaudio
+import torch
+from transformers import pipeline
+
+asr = pipeline("automatic-speech-recognition", model="gigant/romanian-wav2vec2")
+
+def f(audio):
+  rate, sample = audio
+  resampler = torchaudio.transforms.Resample(rate, 16_000)
+  sample_16 = resampler(torch.Tensor(sample)).numpy()
+  return asr(sample_16)["text"]
+
+app = gr.Interface(fn=f, inputs=gr.inputs.Audio(source="upload", type="numpy", label="Audio"), outputs=gr.outputs.Textbox(type="str", label="Predicted text"))
+app.launch()