change to scipy

gradio_asr_en_libri100.py

@@ -13,7 +13,8 @@ import os
 from glob import glob
 from loguru import logger
 # import soundfile as sf
-import librosa
+# import librosa
+from scipy.io import wavfile
 import gradio as gr
 
 from espnet_model_zoo.downloader import ModelDownloader
@@ -53,7 +54,8 @@ logger.info('model loaded')
 def predict(wav_file):
     logger.info('wav file loaded')
     # speech, rate = sf.read(wav_file)
-    speech, rate = librosa.load(wav_file, sr=16000)
+    # speech, rate = librosa.load(wav_file, sr=16000)
+    rate, speech = wavfile.read(wav_file)
     nbests = model(speech)
     text, *_ = nbests[0]
     logger.info('predicted')
@@ -61,16 +63,16 @@ def predict(wav_file):
 
 iface = gr.Interface(
     predict,
-    title='영어 음성인식 데모 (espnet libri100) -- 프로토타입',
-    description='영어 음성 파일을 업로드하면 텍스트 내용을 결과로 보여줍니다.',
+    title='ESPNet2 ASR Librispeech Conformer (trained on clean-100h)',
+    description='Upload your wav file to test the model',
     inputs=[
-        gr.inputs.Audio(label='영어 음성', source='upload', type='filepath')
+        gr.inputs.Audio(label='wav file', source='upload', type='filepath')
     ],
     outputs=[
-        gr.outputs.Textbox(label='음성 인식 디코딩결과'),
+        gr.outputs.Textbox(label='decoding result'),
     ],
     examples=examples,
-    article='<p style="text-align:center">i-Scream AI</p>',
+    article='<p style="text-align:center">Model URL<a href="https://huggingface.co/jkang/espnet2_librispeech_100_conformer">🤗</a></p>',
 )
 
 if __name__ == '__main__':

requirements.txt

@@ -3,4 +3,4 @@ espnet_model_zoo==0.1.7
 gradio==2.7.5.2
 loguru==0.6.0
 librosa
-soundfile
+scipy==1.7.3