Update app.py
Browse files
app.py
CHANGED
@@ -1,53 +1,154 @@
|
|
1 |
-
import gradio as gr
|
2 |
-
import nemo.collections.asr as nemo_asr
|
3 |
-
from pydub import AudioSegment
|
4 |
-
import os
|
5 |
-
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
11 |
-
|
12 |
-
#
|
13 |
-
|
14 |
-
|
15 |
-
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
|
20 |
-
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
|
26 |
-
|
27 |
-
|
28 |
-
|
29 |
-
|
30 |
-
|
31 |
-
|
32 |
-
|
33 |
-
|
34 |
-
|
35 |
-
|
36 |
-
|
37 |
-
|
38 |
-
|
39 |
-
|
40 |
-
|
41 |
-
|
42 |
-
|
43 |
-
|
44 |
-
|
45 |
-
|
46 |
-
|
47 |
-
|
48 |
-
|
49 |
-
|
50 |
-
)
|
51 |
-
|
52 |
-
|
53 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import nemo.collections.asr as nemo_asr
|
3 |
+
from pydub import AudioSegment
|
4 |
+
import os
|
5 |
+
import yt_dlp as youtube_dl
|
6 |
+
from huggingface_hub import login
|
7 |
+
from hazm import Normalizer
|
8 |
+
import numpy as np
|
9 |
+
import re
|
10 |
+
import time
|
11 |
+
|
12 |
+
# Fetch the token from an environment variable
|
13 |
+
HF_TOKEN = os.getenv("HF_TOKEN")
|
14 |
+
if not HF_TOKEN:
|
15 |
+
raise ValueError("HF_TOKEN environment variable not set. Please provide a valid Hugging Face token.")
|
16 |
+
|
17 |
+
# Authenticate with Hugging Face
|
18 |
+
login(HF_TOKEN)
|
19 |
+
|
20 |
+
# Load the private NeMo ASR model
|
21 |
+
try:
|
22 |
+
asr_model = nemo_asr.models.EncDecHybridRNNTCTCBPEModel.from_pretrained(
|
23 |
+
model_name="faimlab/stt_fa_fastconformer_hybrid_large_dataset_v30"
|
24 |
+
)
|
25 |
+
except Exception as e:
|
26 |
+
raise RuntimeError(f"Failed to load model: {str(e)}")
|
27 |
+
|
28 |
+
normalizer = Normalizer()
|
29 |
+
|
30 |
+
def load_audio(audio_path):
|
31 |
+
audio = AudioSegment.from_file(audio_path)
|
32 |
+
audio = audio.set_channels(1).set_frame_rate(16000)
|
33 |
+
audio_samples = np.array(audio.get_array_of_samples(), dtype=np.float32)
|
34 |
+
audio_samples /= np.max(np.abs(audio_samples))
|
35 |
+
return audio_samples, audio.frame_rate
|
36 |
+
|
37 |
+
def transcribe_chunk(audio_chunk, model):
|
38 |
+
transcription = model.transcribe([audio_chunk], batch_size=1, verbose=False)
|
39 |
+
return transcription[0].text
|
40 |
+
|
41 |
+
def transcribe_audio(file_path, model, chunk_size=30*16000):
|
42 |
+
waveform, _ = load_audio(file_path)
|
43 |
+
transcriptions = []
|
44 |
+
for start in range(0, len(waveform), chunk_size):
|
45 |
+
end = min(len(waveform), start + chunk_size)
|
46 |
+
transcription = transcribe_chunk(waveform[start:end], model)
|
47 |
+
transcriptions.append(transcription)
|
48 |
+
|
49 |
+
transcriptions = ' '.join(transcriptions)
|
50 |
+
transcriptions = re.sub(' +', ' ', transcriptions)
|
51 |
+
transcriptions = normalizer.normalize(transcriptions)
|
52 |
+
|
53 |
+
return transcriptions
|
54 |
+
|
55 |
+
# YouTube audio download function
|
56 |
+
YT_LENGTH_LIMIT_S = 3600
|
57 |
+
|
58 |
+
def download_yt_audio(yt_url, filename, cookie_file="cookies.txt"):
|
59 |
+
info_loader = youtube_dl.YoutubeDL()
|
60 |
+
|
61 |
+
try:
|
62 |
+
info = info_loader.extract_info(yt_url, download=False)
|
63 |
+
except youtube_dl.utils.DownloadError as err:
|
64 |
+
raise gr.Error(str(err))
|
65 |
+
|
66 |
+
file_length = info["duration_string"]
|
67 |
+
file_h_m_s = file_length.split(":")
|
68 |
+
file_h_m_s = [int(sub_length) for sub_length in file_h_m_s]
|
69 |
+
|
70 |
+
if len(file_h_m_s) == 1:
|
71 |
+
file_h_m_s.insert(0, 0)
|
72 |
+
if len(file_h_m_s) == 2:
|
73 |
+
file_h_m_s.insert(0, 0)
|
74 |
+
file_length_s = file_h_m_s[0] * 3600 + file_h_m_s[1] * 60 + file_h_m_s[2]
|
75 |
+
|
76 |
+
if file_length_s > YT_LENGTH_LIMIT_S:
|
77 |
+
yt_length_limit_hms = time.strftime("%HH:%MM:%SS", time.gmtime(YT_LENGTH_LIMIT_S))
|
78 |
+
file_length_hms = time.strftime("%HH:%MM:%SS", time.gmtime(file_length_s))
|
79 |
+
raise gr.Error(f"Maximum YouTube length is {yt_length_limit_hms}, got {file_length_hms} YouTube video.")
|
80 |
+
|
81 |
+
ydl_opts = {"outtmpl": filename, "format": "worstvideo[ext=mp4]+bestaudio[ext=m4a]/best[ext=mp4]/best", "cookies": cookie_file}
|
82 |
+
|
83 |
+
with youtube_dl.YoutubeDL(ydl_opts) as ydl:
|
84 |
+
try:
|
85 |
+
ydl.download([yt_url])
|
86 |
+
except youtube_dl.utils.ExtractorError as err:
|
87 |
+
raise gr.Error(str(err))
|
88 |
+
|
89 |
+
|
90 |
+
# Gradio Interface
|
91 |
+
def transcribe(audio):
|
92 |
+
if audio is None:
|
93 |
+
return "Please upload an audio file."
|
94 |
+
|
95 |
+
transcription = transcribe_audio(audio, asr_model)
|
96 |
+
|
97 |
+
return transcription
|
98 |
+
|
99 |
+
def transcribe_yt(yt_url):
|
100 |
+
temp_filename = "/tmp/yt_audio.mp4" # Temporary filename for the downloaded video
|
101 |
+
download_yt_audio(yt_url, temp_filename)
|
102 |
+
transcription = transcribe_audio(temp_filename, asr_model)
|
103 |
+
return transcription
|
104 |
+
|
105 |
+
mf_transcribe = gr.Interface(
|
106 |
+
fn=transcribe,
|
107 |
+
inputs=gr.Microphone(type="filepath"),
|
108 |
+
outputs=gr.Textbox(label="Transcription"),
|
109 |
+
theme="huggingface",
|
110 |
+
title="Persian ASR Transcription with NeMo Fast Conformer",
|
111 |
+
description=(
|
112 |
+
"Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the NeMo's Fast Conformer Hybrid Large.\n\n"
|
113 |
+
"Trained on ~800 hours of Persian speech dataset (Common Voice 17 (~300 hours), YouTube (~400 hours), NasleMana (~90 hours), In-house dataset (~70 hours)).\n\n"
|
114 |
+
"For commercial applications, contact us via email: <[email protected]>.\n\n"
|
115 |
+
"Credit FAIM Group, Sharif University of Technology.\n\n"
|
116 |
+
),
|
117 |
+
allow_flagging="never",
|
118 |
+
)
|
119 |
+
|
120 |
+
# File upload tab
|
121 |
+
file_transcribe = gr.Interface(
|
122 |
+
fn=transcribe,
|
123 |
+
inputs=gr.Audio(type="filepath", label="Audio file"),
|
124 |
+
outputs=gr.Textbox(label="Transcription"),
|
125 |
+
theme="huggingface",
|
126 |
+
title="Persian ASR Transcription with NeMo Fast Conformer",
|
127 |
+
description=(
|
128 |
+
"Transcribe long-form microphone or audio inputs with the click of a button! Demo uses the NeMo's Fast Conformer Hybrid Large.\n\n"
|
129 |
+
"Trained on ~800 hours of Persian speech dataset (Common Voice 17 (~300 hours), YouTube (~400 hours), NasleMana (~90 hours), In-house dataset (~70 hours)).\n\n"
|
130 |
+
"For commercial applications, contact us via email: <[email protected]>.\n\n"
|
131 |
+
"Credit FAIM Group, Sharif University of Technology.\n\n"
|
132 |
+
),
|
133 |
+
allow_flagging="never",
|
134 |
+
)
|
135 |
+
|
136 |
+
# YouTube tab
|
137 |
+
yt_transcribe = gr.Interface(
|
138 |
+
fn=transcribe_yt,
|
139 |
+
inputs=gr.Textbox(label="YouTube URL", placeholder="Enter the YouTube URL here"),
|
140 |
+
outputs=gr.Textbox(label="Transcription"),
|
141 |
+
theme="huggingface",
|
142 |
+
title="Transcribe YouTube Video",
|
143 |
+
description="Transcribe audio from a YouTube video by providing its URL. Currently YouTube is blocking the requests. So you will see the app showing error",
|
144 |
+
allow_flagging="never",
|
145 |
+
)
|
146 |
+
|
147 |
+
# Gradio Interface
|
148 |
+
demo = gr.Blocks()
|
149 |
+
|
150 |
+
with demo:
|
151 |
+
# Create the tabs with the list of interfaces
|
152 |
+
gr.TabbedInterface([mf_transcribe, file_transcribe, yt_transcribe], ["Microphone", "Audio file", "YouTube"])
|
153 |
+
|
154 |
+
demo.launch()
|