DrishtiSharma commited on
Commit
713e80d
1 Parent(s): afeb010

Upload app.py

Browse files
Files changed (1) hide show
  1. app.py +42 -0
app.py ADDED
@@ -0,0 +1,42 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gradio as gr
2
+ import librosa
3
+ from transformers import AutoFeatureExtractor, pipeline
4
+
5
+
6
+ def load_and_fix_data(input_file, model_sampling_rate):
7
+ speech, sample_rate = librosa.load(input_file)
8
+ if len(speech.shape) > 1:
9
+ speech = speech[:, 0] + speech[:, 1]
10
+ if sample_rate != model_sampling_rate:
11
+ speech = librosa.resample(speech, sample_rate, model_sampling_rate)
12
+ return speech
13
+
14
+
15
+ feature_extractor = AutoFeatureExtractor.from_pretrained(
16
+ "anuragshas/wav2vec2-xls-r-1b-hi-with-lm"
17
+ )
18
+ sampling_rate = feature_extractor.sampling_rate
19
+
20
+ asr = pipeline(
21
+ "automatic-speech-recognition", model="anuragshas/wav2vec2-xls-r-1b-hi-with-lm"
22
+ )
23
+
24
+
25
+ def predict_and_ctc_lm_decode(input_file):
26
+ speech = load_and_fix_data(input_file, sampling_rate)
27
+ transcribed_text = asr(speech, chunk_length_s=5, stride_length_s=1)
28
+ return transcribed_text["text"]
29
+
30
+
31
+ gr.Interface(
32
+ predict_and_ctc_lm_decode,
33
+ inputs=[
34
+ gr.inputs.Audio(source="microphone", type="filepath", label="Record your audio")
35
+ ],
36
+ outputs=[gr.outputs.Textbox()],
37
+ examples=[["example1.wav"]],
38
+ title="Hindi ASR using Wav2Vec2-1B with LM",
39
+ description="Built during Robust Speech Event",
40
+ layout="horizontal",
41
+ theme="huggingface",
42
+ ).launch(enable_queue=True)