Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,85 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import speech_recognition as sr
|
3 |
+
from Levenshtein import ratio
|
4 |
+
import tempfile
|
5 |
+
import numpy as np
|
6 |
+
import soundfile as sf
|
7 |
+
import os
|
8 |
+
|
9 |
+
def transcribe_audio(file_info):
|
10 |
+
r = sr.Recognizer()
|
11 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmpfile:
|
12 |
+
sf.write(tmpfile.name, data=file_info[1], samplerate=44100, format='WAV')
|
13 |
+
tmpfile.seek(0)
|
14 |
+
with sr.AudioFile(tmpfile.name) as source:
|
15 |
+
audio_data = r.record(source)
|
16 |
+
os.remove(tmpfile.name) # Cleanup temporary file
|
17 |
+
try:
|
18 |
+
text = r.recognize_google(audio_data)
|
19 |
+
return text
|
20 |
+
except sr.UnknownValueError:
|
21 |
+
return "Could not understand audio"
|
22 |
+
except sr.RequestError as e:
|
23 |
+
return f"Could not request results; {e}"
|
24 |
+
|
25 |
+
def pronunciation_correction(expected_text, file_info):
|
26 |
+
print(f"Expected text: {expected_text}")
|
27 |
+
user_spoken_text = transcribe_audio(file_info)
|
28 |
+
print(f"User spoken text: {user_spoken_text}")
|
29 |
+
|
30 |
+
similarity = ratio(expected_text.lower(), user_spoken_text.lower())
|
31 |
+
description = f"{similarity:.2f}"
|
32 |
+
|
33 |
+
if similarity >= 0.9:
|
34 |
+
feedback = "Excellent pronunciation!"
|
35 |
+
elif similarity >= 0.7:
|
36 |
+
feedback = "Good pronunciation!"
|
37 |
+
elif similarity >= 0.5:
|
38 |
+
feedback = "Needs improvement."
|
39 |
+
else:
|
40 |
+
feedback = "Poor pronunciation, try to focus more on clarity."
|
41 |
+
|
42 |
+
print(f"Similarity: {similarity}, Feedback: {feedback}")
|
43 |
+
return feedback, description
|
44 |
+
|
45 |
+
def validate_sentence(sentence):
|
46 |
+
if not sentence.strip():
|
47 |
+
return "Please enter a sentence."
|
48 |
+
return sentence
|
49 |
+
|
50 |
+
def download_audio(file_info):
|
51 |
+
with tempfile.NamedTemporaryFile(delete=False, suffix=".wav") as tmpfile:
|
52 |
+
sf.write(tmpfile.name, data=file_info[1], samplerate=44100, format='WAV')
|
53 |
+
tmpfile.seek(0)
|
54 |
+
return tmpfile.name # Return file path for download
|
55 |
+
|
56 |
+
with gr.Blocks() as app:
|
57 |
+
with gr.Row():
|
58 |
+
sentence_input = gr.Textbox(label="Enter Your Sentence Here")
|
59 |
+
validated_sentence = gr.Textbox(label="Valid Sentence", interactive=False)
|
60 |
+
audio_input = gr.Audio(label="Upload or Record Audio File", type="numpy")
|
61 |
+
check_pronunciation_button = gr.Button("Check Pronunciation")
|
62 |
+
pronunciation_feedback = gr.Textbox(label="Pronunciation Feedback")
|
63 |
+
pronunciation_score = gr.Number(label="Pronunciation Accuracy Score: 0 (No Match) ~ 1 (Perfect)")
|
64 |
+
download_audio_button = gr.Button("Download Recording")
|
65 |
+
download_output = gr.File(label="Download Your Recording")
|
66 |
+
|
67 |
+
sentence_input.change(
|
68 |
+
validate_sentence,
|
69 |
+
inputs=sentence_input,
|
70 |
+
outputs=validated_sentence
|
71 |
+
)
|
72 |
+
|
73 |
+
check_pronunciation_button.click(
|
74 |
+
pronunciation_correction,
|
75 |
+
inputs=[validated_sentence, audio_input],
|
76 |
+
outputs=[pronunciation_feedback, pronunciation_score]
|
77 |
+
)
|
78 |
+
|
79 |
+
download_audio_button.click(
|
80 |
+
download_audio,
|
81 |
+
inputs=[audio_input],
|
82 |
+
outputs=download_output
|
83 |
+
)
|
84 |
+
|
85 |
+
app.launch(debug=True)
|