[demo] support resampling audio

app.py

@@ -15,7 +15,6 @@
 
 import json
 import gradio as gr
-import numpy as np
 import wenetruntime as wenet
 import librosa
 
@@ -27,12 +26,9 @@ decoder_cn = wenet.Decoder(lang='chs')
 def recognition(audio):
     if audio is None:
         return "Input Error! Please enter one audio!"
-    sr, y = audio
+    y, _ = librosa.load(audio, sr=16000)
     # NOTE: model supports 16k sample_rate
-    if sr != 16000:
-        y = librosa.resample((y / max(np.max(y), 1)).astype(np.float),
-                             sr, 16000)
-        y = (y * (1 << 15)).astype("int16")
+    y = (y * (1 << 15)).astype("int16")
     ans = decoder_cn.decode(y.tobytes(), True)
     if ans is None:
         return "ERROR! No text output! Please try again!"
@@ -46,7 +42,7 @@ def recognition(audio):
 
 
 # input
-inputs = gr.inputs.Audio(source="microphone", type="numpy", label='Input audio')
+inputs = gr.inputs.Audio(source="microphone", type="filepath", label='Input audio')
 
 output = gr.outputs.Textbox(label="Output Text")