Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
@@ -1,11 +1,10 @@
|
|
1 |
-
import librosa
|
2 |
import torch
|
3 |
from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer
|
4 |
import streamlit as st
|
5 |
-
|
6 |
from audio_recorder_streamlit import audio_recorder
|
7 |
|
8 |
-
audio_bytes = audio_recorder()
|
9 |
if audio_bytes:
|
10 |
st.audio(audio_bytes, format="audio/wav")
|
11 |
|
@@ -14,12 +13,14 @@ tokenizer = Wav2Vec2Tokenizer.from_pretrained("facebook/wav2vec2-base-960h")
|
|
14 |
model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-base-960h")
|
15 |
|
16 |
#load audio file
|
17 |
-
speech, rate = librosa.load("/hip-voice.m4a",sr=16000)
|
|
|
|
|
|
|
18 |
|
19 |
-
|
20 |
-
display.Audio("batman1.wav", autoplay=True)
|
21 |
|
22 |
-
input_values = tokenizer(speech, return_tensors = 'pt').input_values
|
23 |
logits = model(input_values).logits
|
24 |
|
25 |
predicted_ids = torch.argmax(logits, dim =-1)
|
@@ -28,5 +29,3 @@ predicted_ids = torch.argmax(logits, dim =-1)
|
|
28 |
transcriptions = tokenizer.decode(predicted_ids[0])
|
29 |
|
30 |
print(transcriptions)
|
31 |
-
|
32 |
-
st.write("hi")
|
|
|
1 |
+
#import librosa
|
2 |
import torch
|
3 |
from transformers import Wav2Vec2ForCTC, Wav2Vec2Tokenizer
|
4 |
import streamlit as st
|
|
|
5 |
from audio_recorder_streamlit import audio_recorder
|
6 |
|
7 |
+
audio_bytes = audio_recorder(pause_threshold=3.0, sample_rate=16_000)
|
8 |
if audio_bytes:
|
9 |
st.audio(audio_bytes, format="audio/wav")
|
10 |
|
|
|
13 |
model = Wav2Vec2ForCTC.from_pretrained("facebook/wav2vec2-base-960h")
|
14 |
|
15 |
#load audio file
|
16 |
+
#speech, rate = librosa.load("/hip-voice.m4a",sr=16000)
|
17 |
+
|
18 |
+
#import IPython.display as display
|
19 |
+
#display.Audio("batman1.wav", autoplay=True)
|
20 |
|
21 |
+
input_values = tokenizer(audio_bytes, return_tensors = 'pt').input_values
|
|
|
22 |
|
23 |
+
#input_values = tokenizer(speech, return_tensors = 'pt').input_values
|
24 |
logits = model(input_values).logits
|
25 |
|
26 |
predicted_ids = torch.argmax(logits, dim =-1)
|
|
|
29 |
transcriptions = tokenizer.decode(predicted_ids[0])
|
30 |
|
31 |
print(transcriptions)
|
|
|
|