[demo] support resampling audio

app.py

@@ -17,6 +17,8 @@ import json
 import gradio as gr
 import numpy as np
 import wenetruntime as wenet
+import librosa
+
 
 wenet.set_log_level(2)
 decoder_cn = wenet.Decoder(lang='chs')
@@ -26,9 +28,9 @@ def recognition(audio):
     if audio is None:
         return "Input Error! Please enter one audio!"
     sr, y = audio
-    assert sr in [48000, 16000]
-    if sr == 48000:  # Optional resample to 16000
-        y = (y / max(np.max(y), 1) * 32767)[::3].astype("int16")
+    # NOTE: model supports 16k sample_rate
+    if sr != 16000:
+        y = librosa.resample(y, sr, 16000)
     ans = decoder_cn.decode(y.tobytes(), True)
     if ans is None:
         return "ERROR! No text output! Please try again!"

requirements.txt

@@ -1,2 +1,3 @@
 wenetruntime
 gradio
+librosa