akshansh36
commited on
Commit
•
cfc3b59
1
Parent(s):
3931c03
Update app.py
Browse files
app.py
CHANGED
@@ -1,27 +1,384 @@
|
|
|
|
1 |
import gradio as gr
|
2 |
-
import
|
|
|
|
|
|
|
3 |
import time
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
4 |
|
5 |
-
def
|
6 |
-
|
7 |
-
|
8 |
-
|
9 |
-
|
10 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
else:
|
12 |
-
|
13 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
|
|
|
|
|
15 |
|
16 |
-
|
17 |
-
|
18 |
-
|
19 |
-
stream = gr.State()
|
20 |
-
clear = gr.Button("Clear")
|
21 |
|
22 |
-
inp.stream(add_to_stream, [inp, stream], [out, stream])
|
23 |
-
clear.click(lambda: [None, None, None], None, [inp, out, stream])
|
24 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
25 |
|
26 |
if __name__ == "__main__":
|
27 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import os
|
2 |
import gradio as gr
|
3 |
+
import spaces
|
4 |
+
from infer_rvc_python import BaseLoader
|
5 |
+
import random
|
6 |
+
import logging
|
7 |
import time
|
8 |
+
import soundfile as sf
|
9 |
+
from infer_rvc_python.main import download_manager
|
10 |
+
import zipfile
|
11 |
+
import edge_tts
|
12 |
+
import asyncio
|
13 |
+
import librosa
|
14 |
+
import traceback
|
15 |
+
import soundfile as sf
|
16 |
+
from pedalboard import Pedalboard, Reverb, Compressor, HighpassFilter
|
17 |
+
from pedalboard.io import AudioFile
|
18 |
+
from pydub import AudioSegment
|
19 |
+
import noisereduce as nr
|
20 |
+
import numpy as np
|
21 |
+
import urllib.request
|
22 |
+
import shutil
|
23 |
+
import threading
|
24 |
+
|
25 |
+
logging.getLogger("infer_rvc_python").setLevel(logging.ERROR)
|
26 |
+
|
27 |
+
# Ensure the correct path to the models directory
|
28 |
+
model_dir = os.path.join(os.path.dirname(__file__), "models")
|
29 |
+
|
30 |
+
converter = BaseLoader(only_cpu=False, hubert_path=None, rmvpe_path=None)
|
31 |
+
|
32 |
+
title = "<center><strong><font size='7'>Vodex AI</font></strong></center>"
|
33 |
+
theme = "aliabid94/new-theme"
|
34 |
+
|
35 |
+
def find_files(directory):
|
36 |
+
file_paths = []
|
37 |
+
for filename in os.listdir(directory):
|
38 |
+
if filename.endswith('.pth') or filename.endswith('.zip') or filename.endswith('.index'):
|
39 |
+
file_paths.append(os.path.join(directory, filename))
|
40 |
+
return file_paths
|
41 |
+
|
42 |
+
def unzip_in_folder(my_zip, my_dir):
|
43 |
+
with zipfile.ZipFile(my_zip) as zip:
|
44 |
+
for zip_info in zip.infolist():
|
45 |
+
if zip_info.is_dir():
|
46 |
+
continue
|
47 |
+
zip_info.filename = os.path.basename(zip_info.filename)
|
48 |
+
zip.extract(zip_info, my_dir)
|
49 |
+
|
50 |
+
def find_my_model(a_, b_):
|
51 |
+
if a_ is None or a_.endswith(".pth"):
|
52 |
+
return a_, b_
|
53 |
+
|
54 |
+
txt_files = []
|
55 |
+
for base_file in [a_, b_]:
|
56 |
+
if base_file is not None and base_file.endswith(".txt"):
|
57 |
+
txt_files.append(base_file)
|
58 |
+
|
59 |
+
directory = os.path.dirname(a_)
|
60 |
+
|
61 |
+
for txt in txt_files:
|
62 |
+
with open(txt, 'r') as file:
|
63 |
+
first_line = file.readline()
|
64 |
+
|
65 |
+
download_manager(
|
66 |
+
url=first_line.strip(),
|
67 |
+
path=directory,
|
68 |
+
extension="",
|
69 |
+
)
|
70 |
+
|
71 |
+
for f in find_files(directory):
|
72 |
+
if f.endswith(".zip"):
|
73 |
+
unzip_in_folder(f, directory)
|
74 |
+
|
75 |
+
model = None
|
76 |
+
index = None
|
77 |
+
end_files = find_files(directory)
|
78 |
+
|
79 |
+
for ff in end_files:
|
80 |
+
if ff.endswith(".pth"):
|
81 |
+
model = os.path.join(directory, ff)
|
82 |
+
gr.Info(f"Model found: {ff}")
|
83 |
+
if ff.endswith(".index"):
|
84 |
+
index = os.path.join(directory, ff)
|
85 |
+
gr.Info(f"Index found: {ff}")
|
86 |
+
|
87 |
+
if not model:
|
88 |
+
gr.Error(f"Model not found in: {end_files}")
|
89 |
+
|
90 |
+
if not index:
|
91 |
+
gr.Warning("Index not found")
|
92 |
+
|
93 |
+
return model, index
|
94 |
+
|
95 |
+
def get_file_size(url):
|
96 |
+
if "huggingface" not in url:
|
97 |
+
raise ValueError("Only downloads from Hugging Face are allowed")
|
98 |
+
|
99 |
+
try:
|
100 |
+
with urllib.request.urlopen(url) as response:
|
101 |
+
info = response.info()
|
102 |
+
content_length = info.get("Content-Length")
|
103 |
+
|
104 |
+
file_size = int(content_length)
|
105 |
+
if file_size > 500000000:
|
106 |
+
raise ValueError("The file is too large. You can only download files up to 500 MB in size.")
|
107 |
+
|
108 |
+
except Exception as e:
|
109 |
+
raise e
|
110 |
+
|
111 |
+
def clear_files(directory):
|
112 |
+
time.sleep(15)
|
113 |
+
print(f"Clearing files: {directory}.")
|
114 |
+
shutil.rmtree(directory)
|
115 |
+
|
116 |
+
def get_my_model(url_data):
|
117 |
+
if not url_data:
|
118 |
+
return None, None
|
119 |
+
|
120 |
+
if "," in url_data:
|
121 |
+
a_, b_ = url_data.split()
|
122 |
+
a_, b_ = a_.strip().replace("/blob/", "/resolve/"), b_.strip().replace("/blob/", "/resolve/")
|
123 |
+
else:
|
124 |
+
a_, b_ = url_data.strip().replace("/blob/", "/resolve/"), None
|
125 |
+
|
126 |
+
out_dir = "downloads"
|
127 |
+
folder_download = str(random.randint(1000, 9999))
|
128 |
+
directory = os.path.join(out_dir, folder_download)
|
129 |
+
os.makedirs(directory, exist_ok=True)
|
130 |
+
|
131 |
+
try:
|
132 |
+
get_file_size(a_)
|
133 |
+
if b_:
|
134 |
+
get_file_size(b_)
|
135 |
+
|
136 |
+
valid_url = [a_] if not b_ else [a_, b_]
|
137 |
+
for link in valid_url:
|
138 |
+
download_manager(
|
139 |
+
url=link,
|
140 |
+
path=directory,
|
141 |
+
extension="",
|
142 |
+
)
|
143 |
+
|
144 |
+
for f in find_files(directory):
|
145 |
+
if f.endswith(".zip"):
|
146 |
+
unzip_in_folder(f, directory)
|
147 |
+
|
148 |
+
model = None
|
149 |
+
index = None
|
150 |
+
end_files = find_files(directory)
|
151 |
+
|
152 |
+
for ff in end_files:
|
153 |
+
if ff.endswith(".pth"):
|
154 |
+
model = ff
|
155 |
+
gr.Info(f"Model found: {ff}")
|
156 |
+
if ff.endswith(".index"):
|
157 |
+
index = ff
|
158 |
+
gr.Info(f"Index found: {ff}")
|
159 |
+
|
160 |
+
if not model:
|
161 |
+
raise ValueError(f"Model not found in: {end_files}")
|
162 |
+
|
163 |
+
if not index:
|
164 |
+
gr.Warning("Index not found")
|
165 |
+
else:
|
166 |
+
index = os.path.abspath(index)
|
167 |
+
|
168 |
+
return os.path.abspath(model), index
|
169 |
+
|
170 |
+
except Exception as e:
|
171 |
+
raise e
|
172 |
+
finally:
|
173 |
+
t = threading.Thread(target=clear_files, args=(directory,))
|
174 |
+
t.start()
|
175 |
+
|
176 |
+
def convert_now(audio_files, random_tag, converter):
|
177 |
+
return converter(
|
178 |
+
audio_files,
|
179 |
+
random_tag,
|
180 |
+
overwrite=False,
|
181 |
+
parallel_workers=8
|
182 |
+
)
|
183 |
|
184 |
+
def apply_noisereduce(audio_list):
|
185 |
+
print("Applying noise reduction")
|
186 |
+
|
187 |
+
result = []
|
188 |
+
for audio_path in audio_list:
|
189 |
+
out_path = f'{os.path.splitext(audio_path)[0]}_noisereduce.wav'
|
190 |
+
|
191 |
+
try:
|
192 |
+
# Load audio file
|
193 |
+
audio = AudioSegment.from_file(audio_path)
|
194 |
+
|
195 |
+
# Convert audio to numpy array
|
196 |
+
samples = np.array(audio.get_array_of_samples())
|
197 |
+
|
198 |
+
reduced_noise = nr.reduce_noise(samples, sr=audio.frame_rate, prop_decrease=0.6)
|
199 |
+
|
200 |
+
|
201 |
+
reduced_audio = AudioSegment(
|
202 |
+
reduced_noise.tobytes(),
|
203 |
+
frame_rate=audio.frame_rate,
|
204 |
+
sample_width=audio.sample_width,
|
205 |
+
channels=audio.channels
|
206 |
+
)
|
207 |
+
|
208 |
+
|
209 |
+
reduced_audio.export(out_path, format="wav")
|
210 |
+
result.append(out_path)
|
211 |
+
|
212 |
+
except Exception as e:
|
213 |
+
traceback.print_exc()
|
214 |
+
print(f"Error in noise reduction: {str(e)}")
|
215 |
+
result.append(audio_path)
|
216 |
+
|
217 |
+
return result
|
218 |
+
|
219 |
+
def run(audio_files, file_m, file_index):
|
220 |
+
if not audio_files:
|
221 |
+
raise ValueError("Please provide an audio file.")
|
222 |
+
|
223 |
+
if isinstance(audio_files, str):
|
224 |
+
audio_files = [audio_files]
|
225 |
+
|
226 |
+
try:
|
227 |
+
duration_base = librosa.get_duration(filename=audio_files[0])
|
228 |
+
print("Duration:", duration_base)
|
229 |
+
except Exception as e:
|
230 |
+
print(e)
|
231 |
+
|
232 |
+
file_m = os.path.join(model_dir, file_m)
|
233 |
+
file_index = os.path.join(model_dir, file_index) if file_index else None
|
234 |
+
|
235 |
+
random_tag = "USER_" + str(random.randint(10000000, 99999999))
|
236 |
+
|
237 |
+
converter.apply_conf(
|
238 |
+
tag=random_tag,
|
239 |
+
file_model=file_m,
|
240 |
+
pitch_algo="rmvpe+",
|
241 |
+
pitch_lvl=0,
|
242 |
+
file_index=file_index,
|
243 |
+
index_influence=0.75,
|
244 |
+
respiration_median_filtering=3,
|
245 |
+
envelope_ratio=0.25,
|
246 |
+
consonant_breath_protection=0.5,
|
247 |
+
resample_sr=44100 if audio_files[0].endswith('.mp3') else 0,
|
248 |
+
)
|
249 |
+
time.sleep(0.1)
|
250 |
+
|
251 |
+
result = convert_now(audio_files, random_tag, converter)
|
252 |
+
result = apply_noisereduce(result)
|
253 |
+
|
254 |
+
return result, result[0] # Assuming result is a list of file paths
|
255 |
+
|
256 |
+
|
257 |
+
def process_audio(audio_file, uploaded_files, file_m, file_index):
|
258 |
+
if audio_file is not None:
|
259 |
+
result, _ = run([audio_file], file_m, file_index)
|
260 |
+
elif uploaded_files is not None:
|
261 |
+
result, _ = run(uploaded_files, file_m, file_index)
|
262 |
+
|
263 |
+
# Create a mapping from filenames to full paths
|
264 |
+
file_mapping = {os.path.basename(path): path for path in result}
|
265 |
+
filenames = list(file_mapping.keys())
|
266 |
+
|
267 |
+
# Return the file_mapping, updated dropdown, initial playback file, and list of all files for download
|
268 |
+
return file_mapping, gr.update(choices=filenames, value=filenames[0], visible=True), file_mapping[filenames[0]], result
|
269 |
+
|
270 |
+
def update_audio_selection(selected_filename, file_mapping):
|
271 |
+
if file_mapping is None:
|
272 |
+
raise ValueError("File mapping is not available.")
|
273 |
+
if selected_filename not in file_mapping:
|
274 |
+
raise ValueError(f"Selected filename {selected_filename} not found in file mapping.")
|
275 |
+
|
276 |
+
# Use the selected filename to find the full path from the mapping
|
277 |
+
full_path = file_mapping[selected_filename]
|
278 |
+
return gr.update(value=full_path, visible=True)
|
279 |
+
|
280 |
+
|
281 |
+
def switch_input(input_type):
|
282 |
+
if input_type == "Record Audio":
|
283 |
+
return gr.update(visible=True), gr.update(visible=False)
|
284 |
else:
|
285 |
+
return gr.update(visible=False), gr.update(visible=True)
|
286 |
+
|
287 |
+
|
288 |
+
def model_conf():
|
289 |
+
model_files = [f for f in os.listdir(model_dir) if f.endswith(".pth")]
|
290 |
+
return gr.Dropdown(
|
291 |
+
label="Select Model File",
|
292 |
+
choices=model_files,
|
293 |
+
value=model_files[0] if model_files else None,
|
294 |
+
interactive=True,
|
295 |
+
)
|
296 |
+
|
297 |
+
def index_conf():
|
298 |
+
index_files = [f for f in os.listdir(model_dir) if f.endswith(".index")]
|
299 |
+
return gr.Dropdown(
|
300 |
+
label="Select Index File",
|
301 |
+
choices=index_files,
|
302 |
+
value=index_files[0] if index_files else None,
|
303 |
+
interactive=True,
|
304 |
+
)
|
305 |
+
|
306 |
+
def audio_conf():
|
307 |
+
return gr.Audio(
|
308 |
+
label="Upload or Record Audio",
|
309 |
+
sources=["upload", "microphone"], # Allows recording via microphone
|
310 |
+
type="filepath",
|
311 |
+
|
312 |
+
)
|
313 |
+
|
314 |
+
def button_conf():
|
315 |
+
return gr.Button(
|
316 |
+
"Inference",
|
317 |
+
variant="primary",
|
318 |
+
)
|
319 |
|
320 |
+
def output_conf():
|
321 |
+
return gr.File(label="Result", file_count="multiple", interactive=False), gr.Audio(label="Play Result",visible=False,show_share_button=False)
|
322 |
|
323 |
+
def get_gui(theme):
|
324 |
+
with gr.Blocks(theme=theme, delete_cache=(3200, 3200)) as app:
|
325 |
+
gr.Markdown(title)
|
|
|
|
|
326 |
|
|
|
|
|
327 |
|
328 |
+
input_type = gr.Radio(["Record Audio", "Upload Files"], label="Select Input Method", value="Record Audio")
|
329 |
+
|
330 |
+
audio = gr.Audio(label="Record Audio", sources="microphone", type="filepath", visible=True)
|
331 |
+
files = gr.File(label="Upload Audio Files", type="filepath", file_count="multiple", visible=False)
|
332 |
+
|
333 |
+
input_type.change(switch_input, inputs=[input_type], outputs=[audio, files])
|
334 |
+
|
335 |
+
model = model_conf()
|
336 |
+
indx = index_conf()
|
337 |
+
button_base = button_conf()
|
338 |
+
|
339 |
+
dropdown = gr.Dropdown(choices=[], label="Select Processed Audio", visible=False)
|
340 |
+
output_audio = gr.Audio(label="Play Selected Audio", visible=False,show_share_button=False)
|
341 |
+
output_files = gr.File(label="Download Processed Audio", file_count="multiple", interactive=False)
|
342 |
+
|
343 |
+
# output_file, output_audio = output_conf()
|
344 |
+
|
345 |
+
file_mapping_state = gr.State()
|
346 |
+
|
347 |
+
button_base.click(
|
348 |
+
process_audio,
|
349 |
+
inputs=[audio, files, model, indx],
|
350 |
+
outputs=[file_mapping_state, dropdown, output_audio, output_files], # Store file mapping in state
|
351 |
+
)
|
352 |
+
|
353 |
+
dropdown.change(
|
354 |
+
update_audio_selection,
|
355 |
+
inputs=[dropdown, file_mapping_state], # Pass the state (file_mapping) and dropdown selection
|
356 |
+
outputs=output_audio, # Play the selected audio file using the full path
|
357 |
+
)
|
358 |
+
|
359 |
+
# gr.Examples(
|
360 |
+
# examples=[
|
361 |
+
# ["./test.ogg", "./model.pth", "./model.index"],
|
362 |
+
# ["./example2/test2.ogg", "./example2/model.pth", "./example2/model.index"],
|
363 |
+
# ],
|
364 |
+
# fn=process_audio,
|
365 |
+
# inputs=[audio, files, model, indx],
|
366 |
+
# outputs=[output_file, output_audio],
|
367 |
+
# cache_examples=False,
|
368 |
+
# )
|
369 |
+
|
370 |
+
return app
|
371 |
|
372 |
if __name__ == "__main__":
|
373 |
+
app = get_gui(theme)
|
374 |
+
app.queue(default_concurrency_limit=40)
|
375 |
+
app.launch(
|
376 |
+
max_threads=40,
|
377 |
+
share=False,
|
378 |
+
show_error=True,
|
379 |
+
quiet=False,
|
380 |
+
debug=False,
|
381 |
+
allowed_paths=["./downloads/"],
|
382 |
+
)
|
383 |
+
|
384 |
+
|