Spaces:
Sleeping
Sleeping
abdullahedcults
commited on
Commit
•
3dd86d3
1
Parent(s):
6557988
Update app.py
Browse files
app.py
CHANGED
@@ -1,12 +1,63 @@
|
|
1 |
import streamlit as st
|
2 |
from audio_recorder_streamlit import audio_recorder
|
|
|
|
|
|
|
3 |
|
4 |
-
|
5 |
-
|
6 |
-
if audio_bytes:
|
7 |
-
st.audio(audio_bytes, format="audio/wav")
|
8 |
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import streamlit as st
|
2 |
from audio_recorder_streamlit import audio_recorder
|
3 |
+
import time
|
4 |
+
import re
|
5 |
+
import os
|
6 |
|
7 |
+
import whisper
|
8 |
+
model = whisper.load_model('medium')
|
|
|
|
|
9 |
|
10 |
+
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
|
11 |
+
|
12 |
+
#loading the tokenizer and the model
|
13 |
+
tokenizer = AutoTokenizer.from_pretrained("Helsinki-NLP/opus-mt-en-hi")
|
14 |
+
model_hindi = AutoModelForSeq2SeqLM.from_pretrained("Helsinki-NLP/opus-mt-en-hi")
|
15 |
+
|
16 |
+
def translator(text):
|
17 |
+
# function to translate English text to Hindi
|
18 |
+
input_ids = tokenizer.encode(text, return_tensors="pt", padding=True)
|
19 |
+
outputs = model_hindi.generate(input_ids)
|
20 |
+
decoded_text = tokenizer.decode(outputs[0], skip_special_tokens=True)
|
21 |
+
return decoded_text
|
22 |
+
|
23 |
+
def split_sentences(generated_text):
|
24 |
+
split_text = re.split(r'(?<!,)[.!?]', generated_text)
|
25 |
+
split_text = [sentence.strip() for sentence in split_text]
|
26 |
+
return split_text
|
27 |
+
|
28 |
+
def transcribe(audio):
|
29 |
+
result = model.transcribe(audio)
|
30 |
+
generated_text = result["text"]
|
31 |
+
|
32 |
+
def process_transcription(generated_text):
|
33 |
+
generated_text = split_sentences(generated_text)
|
34 |
+
processed_text = ""
|
35 |
+
|
36 |
+
for text in generated_text:
|
37 |
+
translated_text = translator(text)
|
38 |
+
processed_text += translated_text + " "
|
39 |
+
|
40 |
+
return processed_text
|
41 |
+
|
42 |
+
text_hindi = process_transcription(generated_text)
|
43 |
+
return result["text"], text_hindi
|
44 |
+
|
45 |
+
def main():
|
46 |
+
st.title("Translate and Transcribe Audio")
|
47 |
+
|
48 |
+
st.write("Click the 'Start Recording' button to start recording your voice. Press 'Stop Recording' when done.")
|
49 |
+
st.write("The transcribed text will be displayed below.")
|
50 |
+
|
51 |
+
audio_bytes = audio_recorder()
|
52 |
+
if audio_bytes:
|
53 |
+
with st.spinner("Transcribing audio... Please wait."):
|
54 |
+
result_text, translated_text = transcribe(audio_bytes)
|
55 |
+
|
56 |
+
st.subheader("Original Text (English):")
|
57 |
+
st.write(result_text)
|
58 |
+
|
59 |
+
st.subheader("Translated Text (Hindi):")
|
60 |
+
st.write(translated_text)
|
61 |
+
|
62 |
+
if __name__ == "__main__":
|
63 |
+
main()
|