test

.gitignore

@@ -0,0 +1,18 @@
+__pycache__/
+weights/
+TEMP/
+logs/
+csvdb/
+
+# Environments
+.env
+.venv
+env/
+venv/
+ENV/
+env.bak/
+venv.bak/
+
+# Models
+hubert_base.pt
+rmvpe.pt

app.py

@@ -1,5 +1,5 @@
 import subprocess, torch, os, traceback, sys, warnings, shutil, numpy as np
-from mega import Mega
+#from mega import Mega
 os.environ["no_proxy"] = "localhost, 127.0.0.1, ::1"
 import threading
 from time import sleep
@@ -160,7 +160,7 @@ def update_fshift_presets(preset, qfrency, tmbre):
 
 i18n = I18nAuto()
 #i18n.print()
-# 判断是否有能用来训练和加速推理的N卡
+# Determine if there are N cards that can be used to train and accelerate inference
 ngpu = torch.cuda.device_count()
 gpu_infos = []
 mem = []
@@ -244,14 +244,8 @@ def load_hubert():
 weight_root = "weights"
 index_root = "logs"
 names = []
-for name in os.listdir(weight_root):
-    if name.endswith(".pth"):
-        names.append(name)
-index_paths = []
-for root, dirs, files in os.walk(index_root, topdown=False):
-    for name in files:
-        if name.endswith(".index") and "trained" not in name:
-            index_paths.append("%s/%s" % (root, name))
+index_paths = ["./logs/joel/added_IVF479_Flat_nprobe_1.index","./logs/jenny/added_IVF533_Flat_nprobe_1.index"]
+file_index=None
 
 
 
@@ -277,13 +271,15 @@ def vc_single(
     f0_up_key = int(f0_up_key)
     try:
         audio = load_audio(input_audio_path, 16000, DoFormant, Quefrency, Timbre)
+        logging.log(logging.INFO, "audio loaded")
         audio_max = np.abs(audio).max() / 0.95
         if audio_max > 1:
             audio /= audio_max
         times = [0, 0, 0]
-        if hubert_model == None:
+        if hubert_model is None:
             load_hubert()
         if_f0 = cpt.get("f0", 1)
+        print(file_index)
         file_index = (
             (
                 file_index.strip(" ")
@@ -293,7 +289,9 @@ def vc_single(
                 .strip(" ")
                 .replace("trained", "added")
             )
-        )  # 防止小白写错，自动帮他替换掉
+        )  # Determine whether there is an N card that can
+        # be used to prevent Xiao Bai from writing incorrectly
+        # and automatically replace and speed up his reasoning
         # file_big_npy = (
         #     file_big_npy.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
         # )
@@ -359,7 +357,8 @@ def vc_multi(
     try:
         dir_path = (
             dir_path.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
-        )  # 防止小白拷路径头尾带了空格和"和回车
+        )  # Uninstall, please select Speaker ID, Timbre, Save Memory, prevent
+        # small white copy, path with spaces at the end and "and enter."
         opt_root = opt_root.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
         os.makedirs(opt_root, exist_ok=True)
         try:
@@ -416,12 +415,14 @@ def vc_multi(
     except:
         yield traceback.format_exc()
 
-# 一个选项卡全局只能有一个音色
-def get_vc(sid):
+# 一A tab can have only one tone globally
+def get_vc(sid, person):
     global n_spk, tgt_sr, net_g, vc, cpt, version
     if sid == "" or sid == []:
         global hubert_model
-        if hubert_model != None:  # 考虑到轮询, 需要加个判断看是否 sid 是由有模型切换到无模型的
+        if hubert_model != None:  # tabs take into account polling,
+            # and you need to add a judgment to see if the SID can
+            # only have one tone to switch from model to no model
             print("clean_empty_cache")
             del net_g, n_spk, vc, hubert_model, tgt_sr  # ,cpt
             hubert_model = net_g = n_spk = vc = hubert_model = tgt_sr = None
@@ -1257,7 +1258,7 @@ def change_choices2():
 audio_files=[]
 for filename in os.listdir("./audios"):
     if filename.endswith(('.wav','.mp3','.ogg','.flac','.m4a','.aac','.mp4')):
-        audio_files.append(os.path.join('./audios',filename).replace('\\', '/'))
+       audio_files.append(os.path.join('./audios',filename).replace('\\', '/'))
         
 def get_index():
     if check_for_name() != '':
@@ -1304,15 +1305,15 @@ def save_to_wav2(dropbox):
     return os.path.join('./audios',os.path.basename(file_path))
     
 def match_index(sid0):
-    folder=sid0.split(".")[0]
-    parent_dir="./logs/"+folder
-    if os.path.exists(parent_dir):
-        for filename in os.listdir(parent_dir):
-            if filename.endswith(".index"):
-                index_path=os.path.join(parent_dir,filename)
-                return index_path
-    else:
-        return ''
+    global file_index
+    if sid0 == '':
+        file_index = ''
+    elif sid0 == "joel.pth":
+        file_index = "./logs/joel/added_IVF479_Flat_nprobe_1.index"
+    elif sid0 == "jenny.pth":
+        file_index = "./logs/jenny/added_IVF533_Flat_nprobe_1.index"
+
+    return file_index
                 
 def check_for_name():
     if len(names) > 0:
@@ -1337,9 +1338,9 @@ def download_from_url(url, model):
     try:
         if "drive.google.com" in url:
             subprocess.run(["gdown", url, "--fuzzy", "-O", zipfile_path])
-        elif "mega.nz" in url:
-            m = Mega()
-            m.download_url(url, './zips')
+        #elif "mega.nz" in url:
+        #    m = Mega()
+        #    m.download_url(url, './zips')
         else:
             subprocess.run(["wget", url, "-O", zipfile_path])
         for filename in os.listdir("./zips"):
@@ -1470,12 +1471,17 @@ def zip_downloader(model):
     else:
         return f'./weights/{model}.pth', "Could not find Index file."
 
-with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
+with gr.Blocks(theme=gr.themes.Base(), title='RVC RULE1 v1') as app:
+    global person
     with gr.Tabs():
         with gr.TabItem("Inference"):
-            gr.HTML("<h1>  RVC V2 Huggingface Version   </h1>")     
-            gr.HTML("<h10>   Huggingface version made by Clebersla   </h10>")   
-            gr.HTML("<h4>  If you want to use this space privately, I recommend you duplicate the space.  </h4>")
+            gr.HTML("<h1>  RVC_RULE1 -- Humans First  </h1>")     
+            gr.HTML('<h3>A few notes about this demo:  Hit "Refresh" if you do not immediately see the model\
+                    choices come up.  Once you have chosen either Joel ("joel.pth") or Jenny ("jenny.pth"),\
+                    make sure that your index setting to the right matches up with the model.  For instance, if Joel is selected\
+                    choose the index file that has joel in the file path.</h3>')
+            gr.HTML("<h10>   Huggingface version v1 -- DT   </h10>")   
+           # gr.HTML("<h4>  If you want to use this space privately, I recommend you duplicate the space.  </h4>")
 
             # Inference Preset Row
             # with gr.Row():
@@ -1487,17 +1493,17 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
 
             # Other RVC stuff
             with gr.Row():
-                sid0 = gr.Dropdown(label="1.Choose your Model.", choices=sorted(names), value=check_for_name())
+                sid0 = gr.Dropdown(label="Choose your Model.", choices=sorted(names), value="joel.pth")
                 refresh_button = gr.Button("Refresh", variant="primary")
                 if check_for_name() != '':
                     get_vc(sorted(names)[0])
                 vc_transform0 = gr.Number(label="Optional: You can change the pitch here or leave it at 0.", value=0)
-                #clean_button = gr.Button(i18n("卸载音色省显存"), variant="primary")
+                #clean_button = gr.Button(i18n("Uninstall the sound saving video memory"), variant="primary")
                 spk_item = gr.Slider(
                     minimum=0,
                     maximum=2333,
                     step=1,
-                    label=i18n("请选择说话人id"),
+                    label=i18n("To uninstall please select Speaker ID Timbre to save the video memory"),
                     value=0,
                     visible=False,
                     interactive=True,
@@ -1526,7 +1532,7 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         refresh_button2 = gr.Button("Refresh", variant="primary", size='sm')
                         record_button.change(fn=save_to_wav, inputs=[record_button], outputs=[input_audio0])
                         record_button.change(fn=change_choices2, inputs=[], outputs=[input_audio0])
-                    with gr.Row():
+                    '''with gr.Row():
                         with gr.Accordion('Text To Speech', open=False):
                             with gr.Column():
                                 lang = gr.Radio(label='Chinese & Japanese do not work with ElevenLabs currently.',choices=['en','es','fr','pt','zh-CN','de','hi','ja'], value='en')
@@ -1549,14 +1555,14 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                                 animation = gr.Video(type='filepath')
                                 refresh_button2.click(fn=change_choices2, inputs=[], outputs=[input_audio0, animation])
                             with gr.Row():
-                                animate_button = gr.Button('Animate')
+                                animate_button = gr.Button('Animate')'''
 
                 with gr.Column():
                     with gr.Accordion("Index Settings", open=False):
                         file_index1 = gr.Dropdown(
                             label="3. Path to your added.index file (if it didn't automatically find it.)",
-                            choices=get_indexes(),
-                            value=get_index(),
+                            choices=[match_index(sid0)] if file_index else match_index("joel.pth"),
+                            value=match_index("joel.pth"),
                             interactive=True,
                             )
                         sid0.change(fn=match_index, inputs=[sid0],outputs=[file_index1])
@@ -1571,7 +1577,7 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         index_rate1 = gr.Slider(
                             minimum=0,
                             maximum=1,
-                            label=i18n("检索特征占比"),
+                            label=i18n("The proportion of features retrieved"),
                             value=0.66,
                             interactive=True,
                             )
@@ -1580,7 +1586,7 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         type='filepath',
                         interactive=False,
                     )
-                    animate_button.click(fn=mouth, inputs=[size, face, vc_output2, faces], outputs=[animation, preview])
+                    #animate_button.click(fn=mouth, inputs=[size, face, vc_output2, faces], outputs=[animation, preview])
                     with gr.Accordion("Advanced Settings", open=False):
                         f0method0 = gr.Radio(
                             label="Optional: Change the Pitch Extraction Algorithm.\nExtraction methods are sorted from 'worst quality' to 'best quality'.\nmangio-crepe may or may not be better than rmvpe in cases where 'smoothness' is more important, but rmvpe is the best overall.",
@@ -1602,7 +1608,8 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         filter_radius0 = gr.Slider(
                             minimum=0,
                             maximum=7,
-                            label=i18n(">=3则使用对harvest音高识别的结果使用中值滤波，数值为滤波半径，使用可以削弱哑音"),
+                            label=i18n(">=3 uses median filtering for the results of Harvest pitch recognition,\
+                                    which is the filter radius, which can weaken the mute"),
                             value=3,
                             step=1,
                             interactive=True,
@@ -1610,7 +1617,7 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         resample_sr0 = gr.Slider(
                             minimum=0,
                             maximum=48000,
-                            label=i18n("后处理重采样至最终采样率，0为不进行重采样"),
+                            label=i18n("Post-process resampling to the final sample rate, 0 is no resampling at all"),
                             value=0,
                             step=1,
                             interactive=True,
@@ -1619,14 +1626,18 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         rms_mix_rate0 = gr.Slider(
                             minimum=0,
                             maximum=1,
-                            label=i18n("输入源音量包络替换输出音量包络融合比例，越靠近1越使用输出包络"),
+                            label=i18n("The input source volume envelope replaces the output\
+                                    volume envelope fusion ratio, and the closer to 1,\
+                                    the output envelope is used"),
                             value=0.21,
                             interactive=True,
                             )
                         protect0 = gr.Slider(
                             minimum=0,
                             maximum=0.5,
-                            label=i18n("保护清辅音和呼吸声，防止电音撕裂等artifact，拉满0.5不开启，调低加大保护力度但可能降低索引效果"),
+                            label=i18n("Protect clean consonants and breathing sounds, prevent artifacts\
+                                    such as electrical tearing, pull 0.5 full and do not turn on,\
+                                    turn down to increase protection but may reduce the index effect"),
                             value=0.33,
                             step=0.01,
                             interactive=True,
@@ -1681,7 +1692,10 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         formant_refresh_button.click(fn=update_fshift_presets,inputs=[formant_preset, qfrency, tmbre],outputs=[formant_preset, qfrency, tmbre])
             with gr.Row():
                 vc_output1 = gr.Textbox("")
-                f0_file = gr.File(label=i18n("F0曲线文件, 可选, 一行一个音高, 代替默认F0及升降调"), visible=False)
+                f0_file = gr.File(label=i18n("Retrieve feature occupancy F0 curve files, optionally,\
+                                            one pitch per line, instead of the default F0 and the upward\
+                                            and downward adjustment stop of the small white copy path with\
+                                            spaces at the beginning and end and carriage return ratio."), visible=False)
                 
                 but0.click(
                     vc_single,
@@ -1708,12 +1722,18 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                 with gr.Row():
                     with gr.Column():
                         vc_transform1 = gr.Number(
-                            label=i18n("变调(整数, 半音数量, 升八度12降八度-12)"), value=0
+                            label=i18n("Pitch inflection\
+                                    (integer, number of chromats,\
+                                    ascending octave 12 descending\
+                                    octave-12)"), value=0
                         )
-                        opt_input = gr.Textbox(label=i18n("指定输出文件夹"), value="opt")
+                        opt_input = gr.Textbox(label=i18n("Specify the output folder"), value="opt")
                         f0method1 = gr.Radio(
                             label=i18n(
-                                "选择音高提取算法,输入歌声可用pm提速,harvest低音好但巨慢无比,crepe效果好但吃GPU"
+                                "Select the pitch extraction algorithm,\
+                                    the input singing voice can be accelerated by pm, the harvest\
+                                    bass is good but extremely slow, and the crepe effect is good\
+                                    but eats the GPU"
                             ),
                             choices=["pm", "harvest", "crepe", "rmvpe"],
                             value="rmvpe",
@@ -1722,19 +1742,22 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         filter_radius1 = gr.Slider(
                             minimum=0,
                             maximum=7,
-                            label=i18n(">=3则使用对harvest音高识别的结果使用中值滤波，数值为滤波半径，使用可以削弱哑音"),
+                            label=i18n(">=3 uses median filtering for the results of Harvest\
+                                    pitch recognition, which is the filter radius, which can weaken the mute"),
                             value=3,
                             step=1,
                             interactive=True,
                         )
                     with gr.Column():
                         file_index3 = gr.Textbox(
-                            label=i18n("特征检索库文件路径,为空则使用下拉的选择结果"),
+                            label=i18n("Feature retrieval library file\
+                                    path, empty to use drop-down\
+                                    selection results"),
                             value="",
                             interactive=True,
                         )
                         file_index4 = gr.Dropdown(
-                            label=i18n("自动检测index路径,下拉式选择(dropdown)"),
+                            label=i18n("Automatic detection of index path, dropdown selection"),
                             choices=sorted(index_paths),
                             interactive=True,
                         )
@@ -1759,7 +1782,8 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         resample_sr1 = gr.Slider(
                             minimum=0,
                             maximum=48000,
-                            label=i18n("后处理重采样至最终采样率，0为不进行重采样"),
+                            label=i18n("Post-process\
+                                    resampling to the final sample rate, 0 is no resampling at all"),
                             value=0,
                             step=1,
                             interactive=True,
@@ -1767,7 +1791,8 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         rms_mix_rate1 = gr.Slider(
                             minimum=0,
                             maximum=1,
-                            label=i18n("输入源音量包络替换输出音量包络融合比例，越靠近1越使用输出包络"),
+                            label=i18n("The input source volume envelope replaces the output volume\
+                                    envelope fusion ratio, and the closer to 1, the output envelope is used"),
                             value=1,
                             interactive=True,
                         )
@@ -1775,7 +1800,10 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                             minimum=0,
                             maximum=0.5,
                             label=i18n(
-                                "保护清辅音和呼吸声，防止电音撕裂等artifact，拉满0.5不开启，调低加大保护力度但可能降低索引效果"
+                                "Protect clean consonants and breathing sounds, prevent\
+                                    artifacts such as electrical tearing, pull 0.5 full\
+                                    and do not turn on, turn down to increase protection\
+                                    but may reduce the index effect"
                             ),
                             value=0.33,
                             step=0.01,
@@ -1783,11 +1811,15 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         )
                     with gr.Column():
                         dir_input = gr.Textbox(
-                            label=i18n("输入待处理音频文件夹路径(去文件管理器地址栏拷就行了)"),
-                            value="E:\codes\py39\\test-20230416b\\todo-songs",
+                            label=i18n("Enter the path to the folder of the audio to be\
+                                    processed (just go to the file manager address bar\
+                                    and copy it.))"),
+                            value="E:\\codes\\py39\\vits_vc_gpu_train\\audio",
                         )
                         inputs = gr.File(
-                            file_count="multiple", label=i18n("也可批量输入音频文件, 二选一, 优先读文件夹")
+                            file_count="multiple", label=i18n("You can also batch input\
+                                        audio files, choose one of the two, and read the\
+                                        folder first")
                         )
                     with gr.Row():
                         format1 = gr.Radio(
@@ -1797,7 +1829,7 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                             interactive=True,
                         )
                         but1 = gr.Button(i18n("转换"), variant="primary")
-                        vc_output3 = gr.Textbox(label=i18n("输出信息"))
+                        vc_output3 = gr.Textbox(label=i18n("Output information"))
                     but1.click(
                         vc_multi,
                         [
@@ -1821,7 +1853,7 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                         [vc_output3],
                     )
                     but1.click(fn=lambda: easy_uploader.clear())
-        with gr.TabItem("Download Model"):
+        '''with gr.TabItem("Download Model"):
             with gr.Row():
                 url=gr.Textbox(label="Enter the URL to the Model:")
             with gr.Row():
@@ -1829,13 +1861,14 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                 download_button=gr.Button("Download")
             with gr.Row():
                 status_bar=gr.Textbox(label="")
-                download_button.click(fn=download_from_url, inputs=[url, model], outputs=[status_bar])
+                download_button.click(fn=download_from_url,\
+                inputs=[url, model], outputs=[status_bar])
             with gr.Row():
                 gr.Markdown(
                 """
                 Made with ❤️ by [Alice Oliveira](https://github.com/aliceoq) | Hosted with ❤️ by [Mateus Elias](https://github.com/mateuseap)
                 """
-                )
+                )'''
 
         def has_two_files_in_pretrained_folder():
             pretrained_folder = "./pretrained/"
@@ -1853,14 +1886,15 @@ with gr.Blocks(theme=gr.themes.Base(), title='Mangio-RVC-Web 💻') as app:
                     with gr.Column():
                         exp_dir1 = gr.Textbox(label="Voice Name:", value="My-Voice")
                         sr2 = gr.Radio(
-                            label=i18n("目标采样率"),
+                            label=i18n("Target sample rate"),
                             choices=["40k", "48k"],
                             value="40k",
                             interactive=True,
                             visible=False
                         )
                         if_f0_3 = gr.Radio(
-                            label=i18n("模型是否带音高指导(唱歌一定要, 语音可以不要)"),
+                            label=i18n("Whether the model has pitch guidance\
+                                    (singing must be, voice can not)"),
                             choices=[True, False],
                             value=True,
                             interactive=True,

gitignore.txt

@@ -4,8 +4,14 @@ TEMP/
 logs/
 csvdb/
 
-# Environment
+# Environments
+.env
+.venv
+env/
 venv/
+ENV/
+env.bak/
+venv.bak/
 
 # Models
 hubert_base.pt

utils.py

@@ -56,7 +56,8 @@ def load_audio(file, sr, DoFormant, Quefrency, Timbre):
         # Requires the ffmpeg CLI and `ffmpeg-python` package to be installed.
         file = (
             file.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
-        )  # 防止小白拷路径头尾带了空格和"和回车
+        )  # The search feature accounts for the "and" and carriage return
+        # ratios that prevent small white copy paths with spaces at the end and ends
         file_formanted = file.strip(" ").strip('"').strip("\n").strip('"').strip(" ")
 
         # print(f"dofor={bool(DoFormant)} timbr={Timbre} quef={Quefrency}\n")