Spaces:
Running
Running
File size: 33,737 Bytes
7b7cc52 25fce83 36a070b 25fce83 0123586 25fce83 7cb0035 7b7cc52 25fce83 7b7cc52 25fce83 7b7cc52 99cadae 36b3e86 0123586 99cadae 7cb0035 9457eae 7cb0035 a4a10af 7cb0035 7b7cc52 25fce83 7b7cc52 25fce83 7b7cc52 25fce83 f94827e 25fce83 f94827e a44f2df 324d1fa a44f2df 879f13a a44f2df 25fce83 a44f2df 25fce83 7b7cc52 25fce83 7b7cc52 25fce83 8215f6a 25fce83 8215f6a 25fce83 6cda5b5 25fce83 06e9a88 25fce83 06e9a88 a44f2df 36a070b 91ba404 25fce83 06e9a88 25fce83 5d1e1a3 25fce83 91ba404 a44f2df 25fce83 7b7cc52 99cadae 515a61a 9f4df50 515a61a 9f4df50 1e10a17 99cadae 515a61a 99cadae 9f4df50 1e10a17 515a61a 99cadae 515a61a 9f4df50 515a61a 9f4df50 515a61a 9f4df50 515a61a 9f4df50 515a61a 1e10a17 515a61a 5ba798f 99cadae 515a61a 99cadae 515a61a 99cadae 515a61a 99cadae 515a61a 9f4df50 1e10a17 63e4fee 9f4df50 63e4fee 9f4df50 63e4fee 9f4df50 63e4fee 9f4df50 7b7cc52 5d1e1a3 1e10a17 5d1e1a3 06e9a88 99cadae 430cc0f 201847d 04d83e7 af384c2 7cb0035 f1714c0 7cb0035 201847d af384c2 04d83e7 247e6dc af384c2 201847d af384c2 a4a10af af384c2 201847d a4a10af 201847d af384c2 25fce83 63e4fee 25fce83 7b7cc52 2ece7cb |
1 2 3 4 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 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695 696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746 747 748 749 750 751 752 753 754 755 756 757 758 759 760 761 762 763 764 765 766 767 768 769 770 771 772 773 774 775 776 777 778 779 780 781 782 783 784 785 786 787 788 789 790 791 792 793 794 795 796 797 798 799 800 801 802 803 804 805 806 807 808 809 810 811 812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875 876 877 878 879 880 881 882 883 884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910 911 912 913 914 915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955 956 957 958 959 960 961 962 |
from typing import Union
from argparse import ArgumentParser
from pathlib import Path
import subprocess
import librosa
import os
import time
import random
import yt_dlp
from search import get_youtube, download_random
import soundfile
import matplotlib.pyplot as plt
import numpy as np
from PIL import Image, ImageDraw, ImageFont
from moviepy.editor import *
from moviepy.video.io.VideoFileClip import VideoFileClip
import moviepy.editor as mpe
import asyncio
import json
import hashlib
from os import path, getenv
from pydub import AudioSegment
import gradio as gr
import torch
import edge_tts
from datetime import datetime
from scipy.io.wavfile import write
import config
import util
from infer_pack.models import (
SynthesizerTrnMs768NSFsid,
SynthesizerTrnMs768NSFsid_nono
)
from vc_infer_pipeline import VC
import tempfile
from openai import OpenAI
def tts(text, model, voice, api_key):
if api_key == '':
raise gr.Error('Please enter your OpenAI API Key')
else:
try:
client = OpenAI(api_key=api_key)
response = client.audio.speech.create(
model=model, # "tts-1","tts-1-hd"
voice=voice, # 'alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'
input=text,
)
except Exception as error:
# Handle any exception that occurs
raise gr.Error("An error occurred while generating speech. Please check your API key and try again.")
print(str(error))
# Create a temp file to save the audio
with tempfile.NamedTemporaryFile(suffix=".mp3", delete=False) as temp_file:
temp_file.write(response.content)
# Get the file path of the temp file
temp_file_path = temp_file.name
return temp_file_path
# music search
def auto_search(name):
save_music_path = '/tmp/downloaded'
if not os.path.exists(save_music_path):
os.makedirs(save_music_path)
config = {'logfilepath': 'musicdl.log', save_music_path: save_music_path, 'search_size_per_source': 5,
'proxies': {}}
save_path = os.path.join(save_music_path, name + '.mp3')
# youtube
get_youtube(name, os.path.join(save_music_path, name))
# task1 = threading.Thread(
# target=get_youtube,
# args=(name, os.path.join(save_music_path, name))
# )
# task1.start()
# task2 = threading.Thread(
# target=download_random,
# args=(name, config, save_path)
# )
# task2.start()
# task1.join(timeout=20)
# task2.join(timeout=10)
if not os.path.exists(save_path):
return "Not Found", None
signal, sampling_rate = soundfile.read(save_path, dtype=np.int16)
# signal, sampling_rate = open_audio(save_path)
return (sampling_rate, signal)
# SadTalker
import os, sys
from src.gradio_demo import SadTalker
try:
import webui # in webui
in_webui = True
except:
in_webui = False
def toggle_audio_file(choice):
if choice == False:
return gr.update(visible=True), gr.update(visible=False)
else:
return gr.update(visible=False), gr.update(visible=True)
def ref_video_fn(path_of_ref_video):
if path_of_ref_video is not None:
return gr.update(value=True)
else:
return gr.update(value=False)
sad_talker = SadTalker("checkpoints", "src/config", lazy_load=True)
# combine video with music
def combine_music(video, audio):
my_clip = mpe.VideoFileClip(video)
audio_background = mpe.AudioFileClip(audio)
final_audio = mpe.CompositeAudioClip([my_clip.audio, audio_background])
final_clip = my_clip.set_audio(final_audio)
final_clip.write_videofile("video.mp4")
return "video.mp4"
# Reference: https://huggingface.co/spaces/zomehwh/rvc-models/blob/main/app.py#L21 # noqa
in_hf_space = getenv('SYSTEM') == 'spaces'
high_quality = True
# Argument parsing
arg_parser = ArgumentParser()
arg_parser.add_argument(
'--hubert',
default=getenv('RVC_HUBERT', 'hubert_base.pt'),
help='path to hubert base model (default: hubert_base.pt)'
)
arg_parser.add_argument(
'--config',
default=getenv('RVC_MULTI_CFG', 'multi_config.json'),
help='path to config file (default: multi_config.json)'
)
arg_parser.add_argument(
'--api',
action='store_true',
help='enable api endpoint'
)
arg_parser.add_argument(
'--cache-examples',
action='store_true',
help='enable example caching, please remember delete gradio_cached_examples folder when example config has been modified' # noqa
)
args = arg_parser.parse_args()
app_css = '''
#model_info img {
max-width: 100px;
max-height: 100px;
float: right;
}
#model_info p {
margin: unset;
}
'''
app = gr.Blocks(
theme=gr.themes.Soft(primary_hue="orange", secondary_hue="slate"),
css=app_css,
analytics_enabled=False
)
# Load hubert model
hubert_model = util.load_hubert_model(config.device, args.hubert)
hubert_model.eval()
# Load models
multi_cfg = json.load(open(args.config, 'r'))
loaded_models = []
for model_name in multi_cfg.get('models'):
print(f'Loading model: {model_name}')
# Load model info
model_info = json.load(
open(path.join('model', model_name, 'config.json'), 'r')
)
# Load RVC checkpoint
cpt = torch.load(
path.join('model', model_name, model_info['model']),
map_location='cpu'
)
tgt_sr = cpt['config'][-1]
cpt['config'][-3] = cpt['weight']['emb_g.weight'].shape[0] # n_spk
if_f0 = cpt.get('f0', 1)
net_g: Union[SynthesizerTrnMs768NSFsid, SynthesizerTrnMs768NSFsid_nono]
if if_f0 == 1:
net_g = SynthesizerTrnMs768NSFsid(
*cpt['config'],
is_half=util.is_half(config.device)
)
else:
net_g = SynthesizerTrnMs768NSFsid_nono(*cpt['config'])
del net_g.enc_q
# According to original code, this thing seems necessary.
print(net_g.load_state_dict(cpt['weight'], strict=False))
net_g.eval().to(config.device)
net_g = net_g.half() if util.is_half(config.device) else net_g.float()
vc = VC(tgt_sr, config)
loaded_models.append(dict(
name=model_name,
metadata=model_info,
vc=vc,
net_g=net_g,
if_f0=if_f0,
target_sr=tgt_sr
))
print(f'Models loaded: {len(loaded_models)}')
# Edge TTS speakers
tts_speakers_list = asyncio.get_event_loop().run_until_complete(edge_tts.list_voices()) # noqa
# Make MV
def make_bars_image(height_values, index, new_height):
# Define the size of the image
width = 512
height = new_height
# Create a new image with a transparent background
image = Image.new('RGBA', (width, height), color=(0, 0, 0, 0))
# Get the image drawing context
draw = ImageDraw.Draw(image)
# Define the rectangle width and spacing
rect_width = 2
spacing = 2
# Define the list of height values for the rectangles
#height_values = [20, 40, 60, 80, 100, 80, 60, 40]
num_bars = len(height_values)
# Calculate the total width of the rectangles and the spacing
total_width = num_bars * rect_width + (num_bars - 1) * spacing
# Calculate the starting position for the first rectangle
start_x = int((width - total_width) / 2)
# Define the buffer size
buffer_size = 80
# Draw the rectangles from left to right
x = start_x
for i, height in enumerate(height_values):
# Define the rectangle coordinates
y0 = buffer_size
y1 = height + buffer_size
x0 = x
x1 = x + rect_width
# Draw the rectangle
draw.rectangle([x0, y0, x1, y1], fill='white')
# Move to the next rectangle position
if i < num_bars - 1:
x += rect_width + spacing
# Rotate the image by 180 degrees
image = image.rotate(180)
# Mirror the image
image = image.transpose(Image.FLIP_LEFT_RIGHT)
# Save the image
image.save('audio_bars_'+ str(index) + '.png')
return 'audio_bars_'+ str(index) + '.png'
def db_to_height(db_value):
# Scale the dB value to a range between 0 and 1
scaled_value = (db_value + 80) / 80
# Convert the scaled value to a height between 0 and 100
height = scaled_value * 50
return height
def infer(title, audio_in, image_in):
# Load the audio file
audio_path = audio_in
audio_data, sr = librosa.load(audio_path)
# Get the duration in seconds
duration = librosa.get_duration(y=audio_data, sr=sr)
# Extract the audio data for the desired time
start_time = 0 # start time in seconds
end_time = duration # end time in seconds
start_index = int(start_time * sr)
end_index = int(end_time * sr)
audio_data = audio_data[start_index:end_index]
# Compute the short-time Fourier transform
hop_length = 512
stft = librosa.stft(audio_data, hop_length=hop_length)
spectrogram = librosa.amplitude_to_db(np.abs(stft), ref=np.max)
# Get the frequency values
freqs = librosa.fft_frequencies(sr=sr, n_fft=stft.shape[0])
# Select the indices of the frequency values that correspond to the desired frequencies
n_freqs = 114
freq_indices = np.linspace(0, len(freqs) - 1, n_freqs, dtype=int)
# Extract the dB values for the desired frequencies
db_values = []
for i in range(spectrogram.shape[1]):
db_values.append(list(zip(freqs[freq_indices], spectrogram[freq_indices, i])))
# Print the dB values for the first time frame
print(db_values[0])
proportional_values = []
for frame in db_values:
proportional_frame = [db_to_height(db) for f, db in frame]
proportional_values.append(proportional_frame)
print(proportional_values[0])
print("AUDIO CHUNK: " + str(len(proportional_values)))
# Open the background image
background_image = Image.open(image_in)
# Resize the image while keeping its aspect ratio
bg_width, bg_height = background_image.size
aspect_ratio = bg_width / bg_height
new_width = 512
new_height = int(new_width / aspect_ratio)
resized_bg = background_image.resize((new_width, new_height))
# Apply black cache for better visibility of the white text
bg_cache = Image.open('black_cache.png')
resized_bg.paste(bg_cache, (0, resized_bg.height - bg_cache.height), mask=bg_cache)
# Create a new ImageDraw object
draw = ImageDraw.Draw(resized_bg)
# Define the text to be added
text = title
font = ImageFont.truetype("NotoSansSC-Regular.otf", 16)
text_color = (255, 255, 255) # white color
# Calculate the position of the text
text_width, text_height = draw.textsize(text, font=font)
x = 30
y = new_height - 70
# Draw the text on the image
draw.text((x, y), text, fill=text_color, font=font)
# Save the resized image
resized_bg.save('resized_background.jpg')
generated_frames = []
for i, frame in enumerate(proportional_values):
bars_img = make_bars_image(frame, i, new_height)
bars_img = Image.open(bars_img)
# Paste the audio bars image on top of the background image
fresh_bg = Image.open('resized_background.jpg')
fresh_bg.paste(bars_img, (0, 0), mask=bars_img)
# Save the image
fresh_bg.save('audio_bars_with_bg' + str(i) + '.jpg')
generated_frames.append('audio_bars_with_bg' + str(i) + '.jpg')
print(generated_frames)
# Create a video clip from the images
clip = ImageSequenceClip(generated_frames, fps=len(generated_frames)/(end_time-start_time))
audio_clip = AudioFileClip(audio_in)
clip = clip.set_audio(audio_clip)
# Set the output codec
codec = 'libx264'
audio_codec = 'aac'
# Save the video to a file
clip.write_videofile("my_video.mp4", codec=codec, audio_codec=audio_codec)
retimed_clip = VideoFileClip("my_video.mp4")
# Set the desired frame rate
new_fps = 25
# Create a new clip with the new frame rate
new_clip = retimed_clip.set_fps(new_fps)
# Save the new clip as a new video file
new_clip.write_videofile("my_video_retimed.mp4", codec=codec, audio_codec=audio_codec)
return "my_video_retimed.mp4"
# mix vocal and non-vocal
def mix(audio1, audio2):
sound1 = AudioSegment.from_file(audio1)
sound2 = AudioSegment.from_file(audio2)
length = len(sound1)
mixed = sound1[:length].overlay(sound2)
mixed.export("song.wav", format="wav")
return "song.wav"
# Bilibili
def youtube_downloader(
video_identifier,
start_time,
end_time,
is_full_song,
output_filename="track.wav",
num_attempts=5,
url_base="",
quiet=False,
force=True,
):
if is_full_song:
ydl_opts = {
'noplaylist': True,
'format': 'bestaudio/best',
'postprocessors': [{
'key': 'FFmpegExtractAudio',
'preferredcodec': 'wav',
}],
"outtmpl": 'dl_audio/youtube_audio',
}
with yt_dlp.YoutubeDL(ydl_opts) as ydl:
ydl.download([video_identifier])
audio_path = "dl_audio/youtube_audio.wav"
return audio_path
else:
output_path = Path(output_filename)
if output_path.exists():
if not force:
return output_path
else:
output_path.unlink()
quiet = "--quiet --no-warnings" if quiet else ""
command = f"""
yt-dlp {quiet} -x --audio-format wav -f bestaudio -o "{output_filename}" --download-sections "*{start_time}-{end_time}" "{url_base}{video_identifier}" # noqa: E501
""".strip()
attempts = 0
while True:
try:
_ = subprocess.check_output(command, shell=True, stderr=subprocess.STDOUT)
except subprocess.CalledProcessError:
attempts += 1
if attempts == num_attempts:
return None
else:
break
if output_path.exists():
return output_path
else:
return None
def audio_separated(audio_input, progress=gr.Progress()):
# start progress
progress(progress=0, desc="Starting...")
time.sleep(0.1)
# check file input
if audio_input is None:
# show progress
for i in progress.tqdm(range(100), desc="Please wait..."):
time.sleep(0.01)
return (None, None, 'Please input audio.')
# create filename
filename = str(random.randint(10000,99999))+datetime.now().strftime("%d%m%Y%H%M%S")
# progress
progress(progress=0.10, desc="Please wait...")
# make dir output
os.makedirs("output", exist_ok=True)
# progress
progress(progress=0.20, desc="Please wait...")
# write
if high_quality:
write(filename+".wav", audio_input[0], audio_input[1])
else:
write(filename+".mp3", audio_input[0], audio_input[1])
# progress
progress(progress=0.50, desc="Please wait...")
# demucs process
if high_quality:
command_demucs = "python3 -m demucs --two-stems=vocals -d cpu "+filename+".wav -o output"
else:
command_demucs = "python3 -m demucs --two-stems=vocals --mp3 --mp3-bitrate 128 -d cpu "+filename+".mp3 -o output"
os.system(command_demucs)
# progress
progress(progress=0.70, desc="Please wait...")
# remove file audio
if high_quality:
command_delete = "rm -v ./"+filename+".wav"
else:
command_delete = "rm -v ./"+filename+".mp3"
os.system(command_delete)
# progress
progress(progress=0.80, desc="Please wait...")
# progress
for i in progress.tqdm(range(80,100), desc="Please wait..."):
time.sleep(0.1)
if high_quality:
return "./output/htdemucs/"+filename+"/vocals.wav","./output/htdemucs/"+filename+"/no_vocals.wav","Successfully..."
else:
return "./output/htdemucs/"+filename+"/vocals.mp3","./output/htdemucs/"+filename+"/no_vocals.mp3","Successfully..."
# https://github.com/fumiama/Retrieval-based-Voice-Conversion-WebUI/blob/main/infer-web.py#L118 # noqa
def vc_func(
input_audio, model_index, pitch_adjust, f0_method, feat_ratio,
filter_radius, rms_mix_rate, resample_option
):
if input_audio is None:
return (None, 'Please provide input audio.')
if model_index is None:
return (None, 'Please select a model.')
model = loaded_models[model_index]
# Reference: so-vits
(audio_samp, audio_npy) = input_audio
# https://huggingface.co/spaces/zomehwh/rvc-models/blob/main/app.py#L49
# Can be change well, we will see
if (audio_npy.shape[0] / audio_samp) > 600 and in_hf_space:
return (None, 'Input audio is longer than 600 secs.')
# Bloody hell: https://stackoverflow.com/questions/26921836/
if audio_npy.dtype != np.float32: # :thonk:
audio_npy = (
audio_npy / np.iinfo(audio_npy.dtype).max
).astype(np.float32)
if len(audio_npy.shape) > 1:
audio_npy = librosa.to_mono(audio_npy.transpose(1, 0))
if audio_samp != 16000:
audio_npy = librosa.resample(
audio_npy,
orig_sr=audio_samp,
target_sr=16000
)
pitch_int = int(pitch_adjust)
resample = (
0 if resample_option == 'Disable resampling'
else int(resample_option)
)
times = [0, 0, 0]
checksum = hashlib.sha512()
checksum.update(audio_npy.tobytes())
output_audio = model['vc'].pipeline(
hubert_model,
model['net_g'],
model['metadata'].get('speaker_id', 0),
audio_npy,
checksum.hexdigest(),
times,
pitch_int,
f0_method,
path.join('model', model['name'], model['metadata']['feat_index']),
feat_ratio,
model['if_f0'],
filter_radius,
model['target_sr'],
resample,
rms_mix_rate,
'v2'
)
out_sr = (
resample if resample >= 16000 and model['target_sr'] != resample
else model['target_sr']
)
print(f'npy: {times[0]}s, f0: {times[1]}s, infer: {times[2]}s')
return ((out_sr, output_audio), 'Success')
async def edge_tts_vc_func(
input_text, model_index, tts_speaker, pitch_adjust, f0_method, feat_ratio,
filter_radius, rms_mix_rate, resample_option
):
if input_text is None:
return (None, 'Please provide TTS text.')
if tts_speaker is None:
return (None, 'Please select TTS speaker.')
if model_index is None:
return (None, 'Please select a model.')
speaker = tts_speakers_list[tts_speaker]['ShortName']
(tts_np, tts_sr) = await util.call_edge_tts(speaker, input_text)
return vc_func(
(tts_sr, tts_np),
model_index,
pitch_adjust,
f0_method,
feat_ratio,
filter_radius,
rms_mix_rate,
resample_option
)
def update_model_info(model_index):
if model_index is None:
return str(
'### Model info\n'
'Please select a model from dropdown above.'
)
model = loaded_models[model_index]
model_icon = model['metadata'].get('icon', '')
return str(
'### Model info\n'
'![model icon]({icon})'
'**{name}**\n\n'
'Author: {author}\n\n'
'Source: {source}\n\n'
'{note}'
).format(
name=model['metadata'].get('name'),
author=model['metadata'].get('author', 'Anonymous'),
source=model['metadata'].get('source', 'Unknown'),
note=model['metadata'].get('note', ''),
icon=(
model_icon
if model_icon.startswith(('http://', 'https://'))
else '/file/model/%s/%s' % (model['name'], model_icon)
)
)
def _example_vc(
input_audio, model_index, pitch_adjust, f0_method, feat_ratio,
filter_radius, rms_mix_rate, resample_option
):
(audio, message) = vc_func(
input_audio, model_index, pitch_adjust, f0_method, feat_ratio,
filter_radius, rms_mix_rate, resample_option
)
return (
audio,
message,
update_model_info(model_index)
)
async def _example_edge_tts(
input_text, model_index, tts_speaker, pitch_adjust, f0_method, feat_ratio,
filter_radius, rms_mix_rate, resample_option
):
(audio, message) = await edge_tts_vc_func(
input_text, model_index, tts_speaker, pitch_adjust, f0_method,
feat_ratio, filter_radius, rms_mix_rate, resample_option
)
return (
audio,
message,
update_model_info(model_index)
)
with app:
gr.HTML("<center>"
"<h1>🥳🎶🎡 - AI歌手数字人+RVC最新算法</h1>"
"</center>")
gr.Markdown("### <center>🌊 - 身临其境般的AI音乐体验,AI歌手“想把我唱给你听”;Powered by [RVC-Project](https://github.com/RVC-Project/Retrieval-based-Voice-Conversion-WebUI)</center>")
gr.Markdown("### <center>更多精彩应用,敬请关注[滔滔AI](http://www.talktalkai.com);滔滔AI,为爱滔滔!💕</center>")
with gr.Tab("🌊 - OpenAI TTS"):
with gr.Row(variant='panel'):
api_key = gr.Textbox(type='password', label='OpenAI API Key', placeholder='请在此填写您的OpenAI API Key')
model = gr.Dropdown(choices=['tts-1','tts-1-hd'], label='请选择模型(tts-1推理更快,tts-1-hd音质更好)', value='tts-1')
voice = gr.Dropdown(choices=['alloy', 'echo', 'fable', 'onyx', 'nova', 'shimmer'], label='请选择一个说话人', value='alloy')
with gr.Row():
with gr.Column():
inp_text = gr.Textbox(label="请填写您想生成的文本(中英文皆可)", placeholder="想说却还没说的 还很多 攒着是因为想写成歌", lines=5)
btn_text = gr.Button("一键开启真实拟声吧", variant="primary")
with gr.Column():
input_audio = gr.Audio(type="filepath", label="OpenAI TTS真实拟声", interactive=False)
btn_text.click(tts, [inp_text, model, voice, api_key], input_audio)
with gr.Tab("🤗 - 轻松提取音乐"):
with gr.Row():
with gr.Column():
ydl_url_input = gr.Textbox(label="音乐视频网址(可直接填写相应的BV号)", value = "https://www.bilibili.com/video/BV...")
with gr.Row():
start = gr.Number(value=0, label="起始时间 (秒)")
end = gr.Number(value=15, label="结束时间 (秒)")
check_full = gr.Checkbox(label="是否上传整首歌曲", info="若勾选则不需要填写起止时间", value=True)
with gr.Accordion('搜索歌曲名上传', open=False):
search_name = gr.Dropdown(label="通过歌曲名搜索", info="选一首您喜欢的歌曲吧", choices=["周杰伦晴天","周杰伦兰亭序","周杰伦七里香","周杰伦花海","周杰伦反方向的钟","周杰伦一路向北","周杰伦稻香","周杰伦明明就","周杰伦爱在西元前","孙燕姿逆光","陈奕迅富士山下","许嵩有何不可","薛之谦其实","邓紫棋光年之外","李荣浩年少有为"])
vc_search = gr.Button("用歌曲名来搜索吧")
ydl_url_submit = gr.Button("提取声音文件吧", variant="primary")
as_audio_submit = gr.Button("去除背景音吧", variant="primary")
with gr.Column():
ydl_audio_output = gr.Audio(label="歌曲原声")
as_audio_input = ydl_audio_output
as_audio_vocals = gr.Audio(label="歌曲人声部分")
as_audio_no_vocals = gr.Audio(label="歌曲伴奏部分", type="filepath")
as_audio_message = gr.Textbox(label="Message", visible=False)
vc_search.click(auto_search, [search_name], [ydl_audio_output])
ydl_url_submit.click(fn=youtube_downloader, inputs=[ydl_url_input, start, end, check_full], outputs=[ydl_audio_output])
as_audio_submit.click(fn=audio_separated, inputs=[as_audio_input], outputs=[as_audio_vocals, as_audio_no_vocals, as_audio_message], show_progress=True, queue=True)
with gr.Row():
with gr.Tab('🎶 - 歌声转换'):
with gr.Row():
with gr.Column():
input_audio = as_audio_vocals
vc_convert_btn = gr.Button('进行歌声转换吧!', variant='primary')
full_song = gr.Button("加入歌曲伴奏吧!", variant="primary")
new_song = gr.Audio(label="AI歌手+伴奏", type="filepath")
pitch_adjust = gr.Slider(
label='变调(默认为0;+2为升高两个key)',
minimum=-12,
maximum=12,
step=1,
value=0
)
f0_method = gr.Radio(
label='人声提取方法(pm时间更短;rmvpe效果更好)',
choices=['pm', 'rmvpe'],
value='pm',
interactive=True
)
with gr.Accordion('更多设置', open=False):
feat_ratio = gr.Slider(
label='Feature ratio',
minimum=0,
maximum=1,
step=0.1,
value=0.6,
visible=False
)
filter_radius = gr.Slider(
label='Filter radius',
minimum=0,
maximum=7,
step=1,
value=3,
visible=False
)
rms_mix_rate = gr.Slider(
label='Volume envelope mix rate',
minimum=0,
maximum=1,
step=0.1,
value=1,
visible=False
)
resample_rate = gr.Dropdown(
[
'Disable resampling',
'16000',
'22050',
'44100',
'48000'
],
label='是否更新采样率(默认为否)',
value='Disable resampling'
)
with gr.Column():
# Model select
model_index = gr.Dropdown(
[
'%s - %s' % (
m['metadata'].get('source', 'Unknown'),
m['metadata'].get('name')
)
for m in loaded_models
],
label='请选择您的AI歌手(必选)',
type='index'
)
# Model info
with gr.Box():
model_info = gr.Markdown(
'### AI歌手信息\n'
'Please select a model from dropdown above.',
elem_id='model_info'
)
output_audio = gr.Audio(label='AI歌手(无伴奏)', type="filepath")
output_msg = gr.Textbox(label='Output message', visible=False)
vc_convert_btn.click(
vc_func,
[
input_audio, model_index, pitch_adjust, f0_method, feat_ratio,
filter_radius, rms_mix_rate, resample_rate
],
[output_audio, output_msg],
api_name='audio_conversion'
)
full_song.click(fn=mix, inputs=[output_audio, as_audio_no_vocals], outputs=[new_song])
model_index.change(
update_model_info,
inputs=[model_index],
outputs=[model_info],
show_progress=False,
queue=False
)
with gr.Tab("📺 - 音乐视频"):
with gr.Row():
with gr.Column():
inp1 = gr.Textbox(label="为视频配上精彩的文案吧(选填)")
inp2 = new_song
inp3 = gr.Image(source='upload', type='filepath', label="上传一张背景图片吧")
btn = gr.Button("生成您的专属音乐视频吧", variant="primary")
with gr.Column():
out1 = gr.Video(label='您的专属音乐视频').style(width=512)
btn.click(fn=infer, inputs=[inp1, inp2, inp3], outputs=[out1])
with gr.Tab("🤵♀️ - AI歌手数字人"):
with gr.Row().style(equal_height=False):
with gr.Column(variant='panel'):
with gr.Tabs(elem_id="sadtalker_source_image"):
with gr.TabItem('图片上传'):
with gr.Row():
source_image = gr.Image(label="请上传一张您喜欢角色的图片", source="upload", type="filepath", elem_id="img2img_image").style(width=512)
with gr.Tabs(elem_id="sadtalker_driven_audio"):
with gr.TabItem('💕倾情演绎'):
with gr.Column(variant='panel'):
driven_audio = output_audio
submit = gr.Button('想把我唱给你听', elem_id="sadtalker_generate", variant='primary')
gen_mv = gr.Button('为视频添加伴奏吧', variant='primary')
with gr.Row():
gen_video = gr.Video(label="AI歌手数字人视频", format="mp4", interactive=False).style(width=256)
inp_mv_1 = gen_video
inp_mv_2 = as_audio_no_vocals
music_video = gr.Video(label="视频+伴奏", format="mp4").style(width=256)
with gr.Column(variant='panel'):
with gr.Tabs(elem_id="sadtalker_checkbox"):
with gr.TabItem('视频设置'):
with gr.Column(variant='panel'):
# width = gr.Slider(minimum=64, elem_id="img2img_width", maximum=2048, step=8, label="Manually Crop Width", value=512) # img2img_width
# height = gr.Slider(minimum=64, elem_id="img2img_height", maximum=2048, step=8, label="Manually Crop Height", value=512) # img2img_width
pose_style = gr.Slider(minimum=0, maximum=46, step=1, label="Pose style", value=0, visible=False) #
size_of_image = gr.Radio([256, 512], value=256, label='face model resolution', info="use 256/512 model?", visible=False) #
preprocess_type = gr.Radio(['crop', 'extfull'], value='crop', label='是否聚焦角色面部', info="crop:视频会聚焦角色面部;extfull:视频会显示图片全貌")
is_still_mode = gr.Checkbox(label="静态模式 (开启静态模式,角色的面部动作会减少;默认开启)", value=True, visible=False)
batch_size = gr.Slider(label="Batch size (数值越大,生成速度越快;若显卡性能好,可增大数值)", step=1, maximum=32, value=4)
enhancer = gr.Checkbox(label="GFPGAN as Face enhancer", visible=False)
submit.click(
fn=sad_talker.test,
inputs=[source_image,
driven_audio,
preprocess_type,
is_still_mode,
enhancer,
batch_size,
size_of_image,
pose_style
],
outputs=[gen_video]
)
gen_mv.click(fn=combine_music, inputs=[inp_mv_1, inp_mv_2], outputs=[music_video])
gr.Markdown("### <center>注意❗:请不要生成会对个人以及组织造成侵害的内容,此程序仅供科研、学习及个人娱乐使用。</center>")
gr.Markdown("<center>🧸 - 如何使用此程序:填写视频网址和视频起止时间后,依次点击“提取声音文件吧”、“去除背景音吧”、“进行歌声转换吧!”、“加入歌曲伴奏吧!”四个按键即可。</center>")
gr.HTML('''
<div class="footer">
<p>🌊🏞️🎶 - 江水东流急,滔滔无尽声。 明·顾璘
</p>
</div>
''')
app.queue(
concurrency_count=1,
max_size=20,
api_open=args.api
).launch(show_error=True) |