Spaces:
Build error
Build error
patrickvonplaten
commited on
Commit
•
dd689bb
1
Parent(s):
0566261
Update app.py
Browse files
app.py
CHANGED
@@ -1,7 +1,49 @@
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
|
|
|
|
2 |
|
3 |
-
def
|
4 |
-
|
5 |
-
|
6 |
-
|
7 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import soundfile as sf
|
2 |
+
import torch
|
3 |
+
from transformers import SpeechEncoderDecoder, Wav2Vec2Processor
|
4 |
import gradio as gr
|
5 |
+
import scipy.signal as sps
|
6 |
+
import sox
|
7 |
|
8 |
+
def convert(inputfile, outfile):
|
9 |
+
sox_tfm = sox.Transformer()
|
10 |
+
sox_tfm.set_output_format(
|
11 |
+
file_type="wav", channels=1, encoding="signed-integer", rate=16000, bits=16
|
12 |
+
)
|
13 |
+
#print(this is not done)
|
14 |
+
sox_tfm.build(inputfile, outfile)
|
15 |
+
|
16 |
+
def read_file(wav):
|
17 |
+
sample_rate, signal = wav
|
18 |
+
signal = signal.mean(-1)
|
19 |
+
number_of_samples = round(len(signal) * float(16000) / sample_rate)
|
20 |
+
resampled_signal = sps.resample(signal, number_of_samples)
|
21 |
+
return resampled_signal
|
22 |
+
|
23 |
+
def parse_transcription(wav_file):
|
24 |
+
'''
|
25 |
+
filename = wav_file.name.split('.')[0]
|
26 |
+
convert(wav_file.name, filename + "16k.wav")
|
27 |
+
speech, _ = sf.read(filename + "16k.wav")
|
28 |
+
|
29 |
+
'''
|
30 |
+
speech = read_file(wav_file)
|
31 |
+
input_values = processor(speech, sampling_rate=16_000, return_tensors="pt").input_values
|
32 |
+
logits = model(input_values).logits
|
33 |
+
predicted_ids = torch.argmax(logits, dim=-1)
|
34 |
+
transcription = processor.decode(predicted_ids[0], skip_special_tokens=True)
|
35 |
+
return transcription
|
36 |
+
|
37 |
+
|
38 |
+
processor = Wav2Vec2Processor.from_pretrained("facebook/wav2vec2-xls-r-300m-en-to-15")
|
39 |
+
model = SpeechEncoderDecoder.from_pretrained("facebook/wav2vec2-xls-r-300m-en-to-15")
|
40 |
+
|
41 |
+
|
42 |
+
processor = Wav2Vec2Processor.from_pretrained("Harveenchadha/vakyansh-wav2vec2-hindi-him-4200")
|
43 |
+
model = Wav2Vec2ForCTC.from_pretrained("Harveenchadha/vakyansh-wav2vec2-hindi-him-4200")
|
44 |
+
|
45 |
+
|
46 |
+
#input_ = gr.inputs.Audio(source="microphone", type="file")
|
47 |
+
input_ = gr.inputs.Audio(source="microphone", type="numpy")
|
48 |
+
gr.Interface(parse_transcription, inputs = input_, outputs="text",
|
49 |
+
analytics_enabled=False, show_tips=False, enable_queue=True).launch(inline=False);
|