|
import gradio as gr |
|
from rvc_remind import * |
|
import wave |
|
import yt_dlp |
|
import shutil, glob |
|
from easyfuncs import download_from_url, CachedModels |
|
os.makedirs("dataset",exist_ok=True) |
|
model_library = CachedModels() |
|
import os |
|
|
|
def get_training_info(audio_file): |
|
if audio_file is None: |
|
return 'Please provide an audio file!' |
|
duration = get_audio_duration(audio_file) |
|
sample_rate = wave.open(audio_file, 'rb').getframerate() |
|
|
|
training_info = { |
|
(0, 2): (150, 'OV2'), |
|
(2, 3): (200, 'OV2'), |
|
(3, 5): (250, 'OV2'), |
|
(5, 10): (300, 'Normal'), |
|
(10, 25): (500, 'Normal'), |
|
(25, 45): (700, 'Normal'), |
|
(45, 60): (1000, 'Normal') |
|
} |
|
|
|
for (min_duration, max_duration), (epochs, pretrain) in training_info.items(): |
|
if min_duration <= duration < max_duration: |
|
break |
|
else: |
|
return 'Duration is not within the specified range!' |
|
|
|
return f'You should use the **{pretrain}** pretrain with **{epochs}** epochs at **{sample_rate/1000}khz** sample rate.' |
|
|
|
|
|
def if_done(done, p): |
|
while 1: |
|
if p.poll() is None: |
|
sleep(0.5) |
|
else: |
|
break |
|
done[0] = True |
|
|
|
def on_button_click(audio_file_path): |
|
return get_training_info(audio_file_path) |
|
|
|
|
|
def downloader_yt(url, save_path, audio_name): |
|
ydl_opts = { |
|
'format': 'bestaudio/best', |
|
'postprocessors': [{ |
|
'key': 'FFmpegExtractAudio', |
|
'preferredcodec': "wav", |
|
}], |
|
'outtmpl': f"audios/{save_path}", |
|
} |
|
|
|
with yt_dlp.YoutubeDL(ydl_opts) as ydl: |
|
ydl.download([url]) |
|
|
|
def download_audio(video_url, save_folder, audio_name): |
|
|
|
save_path = os.path.join(audio_name) |
|
downloader(video_url, save_path, audio_name) |
|
return "Download Complete!" |
|
|
|
|
|
assets_folder = "./assets/pretrained_v2/" |
|
if not os.path.exists(assets_folder): |
|
os.makedirs(assets_folder) |
|
|
|
files = { |
|
|
|
"f0Ov2Super32kD.pth", |
|
"f0Ov2Super40kG.pth", |
|
|
|
|
|
"D-f040k-TITAN-Medium.pth", |
|
"G-f032k-TITAN-Medium.pth", |
|
|
|
|
|
"D_SnowieV3.1_40k.pth", |
|
"G_SnowieV3.1_48k.pth", |
|
|
|
|
|
"RIN_E3_G.pth", |
|
"RIN_E3_D.pth", |
|
|
|
|
|
} |
|
|
|
|
|
theme = gr.themes.Base() |
|
|
|
|
|
|
|
|
|
with gr.Blocks(theme=theme,title="Blane rvc backup") as app: |
|
with gr.Row(): |
|
gr.Markdown( |
|
""" |
|
# Blane RVC |
|
|
|
"Please read Credits!" |
|
""" |
|
) |
|
with gr.Tabs(): |
|
with gr.TabItem("Inference"): |
|
with gr.Row(): |
|
voice_model = gr.Dropdown(label="Model Voice", choices=sorted(names), value=lambda:sorted(names)[0] if len(sorted(names)) > 0 else '', interactive=True) |
|
spk_item = gr.Slider( |
|
minimum=0, |
|
maximum=2333, |
|
step=1, |
|
label="Speaker ID", |
|
value=0, |
|
|
|
interactive=True, |
|
) |
|
vc_transform0 = gr.Number( |
|
label="Pitch", |
|
value=0 |
|
) |
|
refresh_button = gr.Button("Refresh", variant="primary") |
|
but0 = gr.Button(value="Convert", variant="primary") |
|
with gr.Row(): |
|
with gr.Column(): |
|
with gr.Row(): |
|
dropbox = gr.File(label="Drop your audio here & hit the Reload button.") |
|
with gr.Row(): |
|
paths_for_files = lambda path:[os.path.abspath(os.path.join(path, f)) for f in os.listdir(path) if os.path.splitext(f)[1].lower() in ('.mp3', '.wav', '.flac', '.ogg')] |
|
input_audio0 = gr.Dropdown( |
|
label="Input Path", |
|
value=paths_for_files('audios')[0] if len(paths_for_files('audios')) > 0 else '', |
|
choices=paths_for_files('audios'), |
|
allow_custom_value=True |
|
) |
|
with gr.Row(): |
|
audio_player = gr.Audio() |
|
input_audio0.change( |
|
inputs=[input_audio0], |
|
outputs=[audio_player], |
|
fn=lambda path: {"value":path,"__type__":"update"} if os.path.exists(path) else None |
|
) |
|
dropbox.upload( |
|
fn=lambda audio:audio.name, |
|
inputs=[dropbox], |
|
outputs=[input_audio0]) |
|
with gr.Column(): |
|
with gr.Accordion("Change Index", open=False): |
|
file_index2 = gr.Dropdown( |
|
label="Change Index", |
|
choices=sorted(index_paths), |
|
interactive=True, |
|
value=sorted(index_paths)[0] if len(sorted(index_paths)) > 0 else '' |
|
) |
|
index_rate1 = gr.Slider( |
|
minimum=0, |
|
maximum=1, |
|
label="Index Strength", |
|
value=0.5, |
|
interactive=True, |
|
) |
|
vc_output2 = gr.Audio(label="Output") |
|
with gr.Accordion("General Settings", open=False): |
|
f0method0 = gr.Radio( |
|
label="Method", |
|
choices=["pm", "harvest", "crepe", "rmvpe"], |
|
value="rmvpe", |
|
interactive=True, |
|
) |
|
filter_radius0 = gr.Slider( |
|
minimum=0, |
|
maximum=7, |
|
label="Breathiness Reduction (Harvest only)", |
|
value=3, |
|
step=1, |
|
interactive=True, |
|
) |
|
resample_sr0 = gr.Slider( |
|
minimum=0, |
|
maximum=48000, |
|
label="Resample", |
|
value=0, |
|
step=1, |
|
interactive=True, |
|
visible=False |
|
) |
|
rms_mix_rate0 = gr.Slider( |
|
minimum=0, |
|
maximum=1, |
|
label="Volume Normalization", |
|
value=0, |
|
interactive=True, |
|
) |
|
protect0 = gr.Slider( |
|
minimum=0, |
|
maximum=0.5, |
|
label="Breathiness Protection (0 is enabled, 0.5 is disabled)", |
|
value=0.33, |
|
step=0.01, |
|
interactive=True, |
|
) |
|
if voice_model != None: vc.get_vc(voice_model.value,protect0,protect0) |
|
file_index1 = gr.Textbox( |
|
label="Index Path", |
|
interactive=True, |
|
visible=False |
|
) |
|
refresh_button.click( |
|
fn=change_choices, |
|
inputs=[], |
|
outputs=[voice_model, file_index2], |
|
api_name="infer_refresh", |
|
) |
|
refresh_button.click( |
|
fn=lambda:{"choices":paths_for_files('audios'),"__type__":"update"}, |
|
inputs=[], |
|
outputs = [input_audio0], |
|
) |
|
refresh_button.click( |
|
fn=lambda:{"value":paths_for_files('audios')[0],"__type__":"update"} if len(paths_for_files('audios')) > 0 else {"value":"","__type__":"update"}, |
|
inputs=[], |
|
outputs = [input_audio0], |
|
) |
|
with gr.Row(): |
|
f0_file = gr.File(label="F0 Path", visible=False) |
|
with gr.Row(): |
|
vc_output2 = gr.Audio(label="Information") |
|
with gr.Row(): |
|
vc_output1 = gr.Textbox(label="Information") |
|
but0.click( |
|
vc.vc_single, |
|
[ |
|
spk_item, |
|
input_audio0, |
|
vc_transform0, |
|
f0_file, |
|
f0method0, |
|
file_index1, |
|
file_index2, |
|
index_rate1, |
|
filter_radius0, |
|
resample_sr0, |
|
rms_mix_rate0, |
|
protect0, |
|
], |
|
[vc_output1, vc_output2], |
|
api_name="infer_convert", |
|
) |
|
voice_model.change( |
|
fn=vc.get_vc, |
|
inputs=[voice_model, protect0, protect0], |
|
outputs=[spk_item, protect0, protect0, file_index2, file_index2], |
|
api_name="infer_change_voice", |
|
) |
|
with gr.TabItem("Download Models"): |
|
with gr.Row(): |
|
url_input = gr.Textbox(label="URL to model", value="",placeholder="https://...", scale=6) |
|
name_output = gr.Textbox(label="Save as", value="",placeholder="MyModel",scale=2) |
|
url_download = gr.Button(value="Download Model",scale=2) |
|
url_download.click( |
|
inputs=[url_input,name_output], |
|
outputs=[url_input], |
|
fn=download_from_url, |
|
) |
|
with gr.Row(): |
|
model_browser = gr.Dropdown(choices=list(model_library.models.keys()),label="OR Search Models (Quality UNKNOWN)",scale=5) |
|
download_from_browser = gr.Button(value="Get Voice models",scale=2) |
|
download_from_browser.click( |
|
inputs=[model_browser], |
|
outputs=[model_browser], |
|
fn=lambda model: download_from_url(model_library.models[model],model), |
|
) |
|
with gr.TabItem(" vocal separation"): |
|
with gr.Group(): |
|
gr.Markdown( |
|
value=( |
|
"Batch processing of vocal accompaniment separation using the UVR5 model. <br>Example of a valid folder path format: E:\\codes\\py39\\vits_vc_gpu\\White Egret Frost Testing Samples (just copy from the file manager address bar). <br>The model is divided into three categories: <br>1. Preserve vocals: Choose this for audio without background vocals, it preserves the main vocals better than HP5. It includes two built-in models, HP2 and HP3. HP3 may slightly leak accompaniment but preserves the main vocals slightly better than HP2; <br>2. Only preserve main vocals: Choose this for audio with background vocals, it may weaken the main vocals. It includes one built-in model, HP5; <br>3. Dereverberation and delay removal model (by FoxJoy):<br>(1)MDX-Net(onnx_dereverb): The best choice for dual-channel reverberation, cannot remove single-channel reverberation:" |
|
) |
|
) |
|
with gr.Row(): |
|
with gr.Column(): |
|
dir_wav_input = gr.Textbox( |
|
label=("Enter the path of the audio folder to be processed"), |
|
placeholder="C:\\Users\\Desktop\\todo-songs", |
|
) |
|
wav_inputs = gr.File( |
|
file_count="multiple", |
|
label=("You can also input audio files in batches, choose one of the two options, and prioritize reading the folder."), |
|
) |
|
with gr.Column(): |
|
model_choose = gr.Dropdown( |
|
label=("choices"), choices=uvr5_names |
|
) |
|
agg = gr.Slider( |
|
minimum=0, |
|
maximum=20, |
|
step=1, |
|
label="Vocal extraction aggressiveness", |
|
value=10, |
|
interactive=True, |
|
visible=False, |
|
) |
|
opt_vocal_root = gr.Textbox( |
|
label=("Specify the output folder for the lead vocals"), value="opt" |
|
) |
|
opt_ins_root = gr.Textbox( |
|
label=("Specify the folder for outputting non-lead vocals"), value="opt" |
|
) |
|
format0 = gr.Radio( |
|
label=i18n("Export file formats"), |
|
choices=["wav", "flac", "mp3"], |
|
value="wav", |
|
interactive=True, |
|
) |
|
but2uvr5 = gr.Button("Conversion", variant="primary") |
|
with gr.Row(): |
|
uvrc_output4 = gr.Textbox(label="Output information") |
|
but2uvr5.click( |
|
uvr, |
|
[ |
|
model_choose, |
|
dir_wav_input, |
|
opt_vocal_root, |
|
wav_inputs, |
|
opt_ins_root, |
|
agg, |
|
format0, |
|
], |
|
[uvrc_output4], |
|
api_name="uvr_convert", |
|
) |
|
|
|
with gr.TabItem("Train"): |
|
with gr.Row(): |
|
with gr.Column(): |
|
training_name = gr.Textbox(label="Name your model", value="My-Voice",placeholder="My-Voice") |
|
np7 = gr.Slider( |
|
minimum=0, |
|
maximum=config.n_cpu, |
|
step=1, |
|
label="Number of CPU processes used to extract pitch features", |
|
value=int(np.ceil(config.n_cpu / 1.5)), |
|
interactive=True, |
|
) |
|
sr2 = gr.Radio( |
|
label="Sampling Rate", |
|
choices=["40k", "32k"], |
|
value="32k", |
|
interactive=True, |
|
visible=False |
|
) |
|
if_f0_3 = gr.Radio( |
|
label="Will your model be used for singing? If not, you can ignore this.", |
|
choices=[True, False], |
|
value=True, |
|
interactive=True, |
|
visible=False |
|
) |
|
version19 = gr.Radio( |
|
label="Version", |
|
choices=["v1", "v2"], |
|
value="v2", |
|
interactive=True, |
|
visible=False, |
|
) |
|
dataset_folder = gr.Textbox( |
|
label="dataset folder", value='dataset' |
|
) |
|
easy_uploader = gr.Files(label="Drop your audio files here",file_types=['audio']) |
|
but1 = gr.Button("1. Process", variant="primary") |
|
info1 = gr.Textbox(label="Information", value="",visible=True) |
|
easy_uploader.upload(inputs=[dataset_folder],outputs=[],fn=lambda folder:os.makedirs(folder,exist_ok=True)) |
|
easy_uploader.upload( |
|
fn=lambda files,folder: [shutil.copy2(f.name,os.path.join(folder,os.path.split(f.name)[1])) for f in files] if folder != "" else gr.Warning('Please enter a folder name for your dataset'), |
|
inputs=[easy_uploader, dataset_folder], |
|
outputs=[]) |
|
gpus6 = gr.Textbox( |
|
label="Enter the GPU numbers to use separated by -, (e.g. 0-1-2)", |
|
value=gpus, |
|
interactive=True, |
|
visible=F0GPUVisible, |
|
) |
|
gpu_info9 = gr.Textbox( |
|
label="GPU Info", value=gpu_info, visible=F0GPUVisible |
|
) |
|
spk_id5 = gr.Slider( |
|
minimum=0, |
|
maximum=4, |
|
step=1, |
|
label="Speaker ID", |
|
value=0, |
|
interactive=True, |
|
visible=False |
|
) |
|
but1.click( |
|
preprocess_dataset, |
|
[dataset_folder, training_name, sr2, np7], |
|
[info1], |
|
api_name="train_preprocess", |
|
) |
|
with gr.Column(): |
|
f0method8 = gr.Radio( |
|
label="F0 extraction method", |
|
choices=["pm", "harvest", "dio", "rmvpe", "rmvpe_gpu"], |
|
value="rmvpe_gpu", |
|
interactive=True, |
|
) |
|
gpus_rmvpe = gr.Textbox( |
|
label="GPU numbers to use separated by -, (e.g. 0-1-2)", |
|
value="%s-%s" % (gpus, gpus), |
|
interactive=True, |
|
visible=F0GPUVisible, |
|
) |
|
but2 = gr.Button("2. Extract Features", variant="primary") |
|
info2 = gr.Textbox(label="Information", value="", max_lines=8) |
|
f0method8.change( |
|
fn=change_f0_method, |
|
inputs=[f0method8], |
|
outputs=[gpus_rmvpe], |
|
) |
|
but2.click( |
|
extract_f0_feature, |
|
[ |
|
gpus6, |
|
np7, |
|
f0method8, |
|
if_f0_3, |
|
training_name, |
|
version19, |
|
gpus_rmvpe, |
|
], |
|
[info2], |
|
api_name="train_extract_f0_feature", |
|
) |
|
with gr.Column(): |
|
total_epoch11 = gr.Slider( |
|
minimum=2, |
|
maximum=1000, |
|
step=1, |
|
label="Epochs (more epochs may improve quality but takes longer)", |
|
value=150, |
|
interactive=True, |
|
) |
|
but4 = gr.Button("3. Train Index", variant="primary") |
|
but3 = gr.Button("4. Train Model", variant="primary") |
|
info3 = gr.Textbox(label="Information", value="", max_lines=10) |
|
with gr.Accordion(label="Change pretrains", open=False): |
|
pretrained = lambda sr, letter: [os.path.abspath(os.path.join('assets/pretrained_v2', file)) for file in os.listdir('assets/pretrained_v2') if file.endswith('.pth') and sr in file and letter in file] |
|
pretrained_G14 = gr.Dropdown( |
|
label="pretrained G", |
|
choices=pretrained(sr2.value, 'G'), |
|
value=pretrained(sr2.value, 'G')[0] if len(pretrained(sr2.value, 'G')) > 0 else '', |
|
interactive=True, |
|
visible=True |
|
) |
|
pretrained_D15 = gr.Dropdown( |
|
label="pretrained D", |
|
choices=pretrained(sr2.value, 'D'), |
|
value=pretrained(sr2.value, 'D')[0] if len(pretrained(sr2.value, 'G')) > 0 else '', |
|
visible=True, |
|
interactive=True |
|
) |
|
|
|
with gr.Accordion(label="General Settings", open=False): |
|
gpus16 = gr.Textbox( |
|
label="GPUs separated by -, (e.g. 0-1-2)", |
|
value="0", |
|
interactive=True, |
|
visible=True |
|
) |
|
save_epoch10 = gr.Slider( |
|
minimum=1, |
|
maximum=50, |
|
step=1, |
|
label="Weight Saving Frequency", |
|
value=25, |
|
interactive=True, |
|
) |
|
batch_size12 = gr.Slider( |
|
minimum=1, |
|
maximum=40, |
|
step=1, |
|
label="Batch Size", |
|
value=default_batch_size, |
|
interactive=True, |
|
) |
|
if_save_latest13 = gr.Radio( |
|
label="Only save the latest model", |
|
choices=["yes", "no"], |
|
value="yes", |
|
interactive=True, |
|
visible=False |
|
) |
|
if_cache_gpu17 = gr.Radio( |
|
label="If your dataset is UNDER 10 minutes, cache it to train faster", |
|
choices=["yes", "no"], |
|
value="no", |
|
interactive=True, |
|
) |
|
if_save_every_weights18 = gr.Radio( |
|
label="Save small model at every save point", |
|
choices=["yes", "no"], |
|
value="yes", |
|
interactive=True, |
|
) |
|
with gr.Row(): |
|
download_model = gr.Button('5.Download Model') |
|
with gr.Row(): |
|
model_files = gr.Files(label='Your Model and Index file can be downloaded here:') |
|
download_model.click( |
|
fn=lambda name: os.listdir(f'assets/weights/{name}') + glob.glob(f'logs/{name.split(".")[0]}/added_*.index'), |
|
inputs=[training_name], |
|
outputs=[model_files, info3]) |
|
with gr.Row(): |
|
sr2.change( |
|
change_sr2, |
|
[sr2, if_f0_3, version19], |
|
[pretrained_G14, pretrained_D15], |
|
) |
|
version19.change( |
|
change_version19, |
|
[sr2, if_f0_3, version19], |
|
[pretrained_G14, pretrained_D15, sr2], |
|
) |
|
if_f0_3.change( |
|
change_f0, |
|
[if_f0_3, sr2, version19], |
|
[f0method8, pretrained_G14, pretrained_D15], |
|
) |
|
with gr.Row(): |
|
but5 = gr.Button("1 Click Training", variant="primary", visible=True) |
|
but3.click( |
|
click_train, |
|
[ |
|
training_name, |
|
sr2, |
|
if_f0_3, |
|
spk_id5, |
|
save_epoch10, |
|
total_epoch11, |
|
batch_size12, |
|
if_save_latest13, |
|
pretrained_G14, |
|
pretrained_D15, |
|
gpus16, |
|
if_cache_gpu17, |
|
if_save_every_weights18, |
|
version19, |
|
], |
|
info3, |
|
api_name="train_start", |
|
) |
|
but4.click(train_index, [training_name, version19], info3) |
|
but5.click( |
|
train1key, |
|
[ |
|
training_name, |
|
sr2, |
|
if_f0_3, |
|
dataset_folder, |
|
spk_id5, |
|
np7, |
|
f0method8, |
|
save_epoch10, |
|
total_epoch11, |
|
batch_size12, |
|
if_save_latest13, |
|
pretrained_G14, |
|
pretrained_D15, |
|
gpus16, |
|
if_cache_gpu17, |
|
if_save_every_weights18, |
|
version19, |
|
gpus_rmvpe, |
|
], |
|
info3, |
|
api_name="train_start_all", |
|
) |
|
with gr.TabItem("Extra"): |
|
with gr.Accordion('Training Helper', open=True): |
|
input_audio = gr.Audio(type="filepath", label="Upload your audio file") |
|
gr.Textbox("Please note that these results are approximate and intended to provide a general idea for beginners.", label='Notice:') |
|
training_info_text = gr.Markdown(label="Training Information:") |
|
get_info_button = gr.Button("Get Training Info") |
|
get_info_button.click( |
|
fn=on_button_click, |
|
inputs=[input_audio], |
|
outputs=[training_info_text] |
|
) |
|
with gr.Accordion('audio downloader for inference (vocal only for this one!)', open=True): |
|
url = gr.Textbox(label="input url") |
|
audio_name = gr.Textbox(label="audio name for audio (no space)") |
|
output_songs = gr.Audio(label="outputs songs") |
|
get_info_button = gr.Button("Get audios") |
|
get_info_button.click( |
|
fn=downloader_yt, |
|
inputs=[url, audio_name], |
|
outputs=[output_songs] |
|
) |
|
|
|
with gr.TabItem('credits'): |
|
with gr.Accordion('Credits', open=True): |
|
gr.Markdown( |
|
''' |
|
# CREDITS |
|
|
|
- **RVC Project**: Original Developers |
|
- **blane187**: colab + GUI mod |
|
- **rejetks**: Original EASY GUI Coder |
|
- **eddycrack864**: Helper |
|
|
|
|
|
''') |
|
|
|
if config.iscolab: |
|
app.launch(share=True) |
|
else: |
|
app.launch( |
|
server_name="0.0.0.0", |
|
inbrowser=not config.noautoopen, |
|
server_port=config.listen_port, |
|
quiet=True, |
|
) |