Spaces:
Sleeping
Sleeping
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,70 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
|
2 |
+
import os
|
3 |
+
import uuid
|
4 |
+
import torch
|
5 |
+
import requests
|
6 |
+
import gradio as gr
|
7 |
+
from moviepy.editor import VideoFileClip
|
8 |
+
from speechbrain.pretrained.interfaces import foreign_class
|
9 |
+
|
10 |
+
# Load the pretrained model
|
11 |
+
classifier = foreign_class(
|
12 |
+
source="Jzuluaga/accent-id-commonaccent_xlsr-en-english",
|
13 |
+
pymodule_file="custom_interface.py",
|
14 |
+
classname="CustomEncoderWav2vec2Classifier"
|
15 |
+
)
|
16 |
+
|
17 |
+
def extract_audio(video_path, output_wav="output.wav"):
|
18 |
+
video = VideoFileClip(video_path)
|
19 |
+
audio = video.audio
|
20 |
+
audio.write_audiofile(output_wav, codec='pcm_s16le', fps=16000)
|
21 |
+
return output_wav
|
22 |
+
|
23 |
+
def download_video(url, filename="temp.mp4"):
|
24 |
+
response = requests.get(url, stream=True)
|
25 |
+
with open(filename, 'wb') as f:
|
26 |
+
for chunk in response.iter_content(chunk_size=8192):
|
27 |
+
f.write(chunk)
|
28 |
+
return filename
|
29 |
+
|
30 |
+
def classify_video_accent(video_url):
|
31 |
+
uid = str(uuid.uuid4())
|
32 |
+
video_path = f"{uid}.mp4"
|
33 |
+
wav_path = f"{uid}.wav"
|
34 |
+
|
35 |
+
try:
|
36 |
+
download_video(video_url, video_path)
|
37 |
+
extract_audio(video_path, wav_path)
|
38 |
+
|
39 |
+
out_prob, score, index, text_lab = classifier.classify_file(wav_path)
|
40 |
+
confidence = torch.max(out_prob).item() * 100
|
41 |
+
|
42 |
+
return {
|
43 |
+
"accent": text_lab,
|
44 |
+
"confidence_score": f"{confidence:.2f}%",
|
45 |
+
"summary": f"The speaker is most likely using a(n) {text_lab} English accent."
|
46 |
+
}
|
47 |
+
finally:
|
48 |
+
for f in [video_path, wav_path]:
|
49 |
+
if os.path.exists(f):
|
50 |
+
os.remove(f)
|
51 |
+
|
52 |
+
def gradio_accent_classifier(video_url):
|
53 |
+
try:
|
54 |
+
result = classify_video_accent(video_url)
|
55 |
+
return f"Accent: {result['accent']}
|
56 |
+
Confidence: {result['confidence_score']}
|
57 |
+
Summary: {result['summary']}"
|
58 |
+
except Exception as e:
|
59 |
+
return f"Error: {str(e)}"
|
60 |
+
|
61 |
+
iface = gr.Interface(
|
62 |
+
fn=gradio_accent_classifier,
|
63 |
+
inputs=gr.Textbox(label="Public .mp4 Video URL"),
|
64 |
+
outputs="text",
|
65 |
+
title="English Accent Classifier",
|
66 |
+
description="Paste a direct link to a public .mp4 file to classify the English accent spoken in the video."
|
67 |
+
)
|
68 |
+
|
69 |
+
if __name__ == "__main__":
|
70 |
+
iface.launch()
|