Upload 4 files

app.py

@@ -0,0 +1,676 @@
+import os
+import glob
+import json
+import traceback
+import logging
+import gradio as gr
+import numpy as np
+import librosa
+import torch
+import asyncio
+import edge_tts
+import yt_dlp
+import ffmpeg
+import subprocess
+import sys
+import io
+import wave
+from datetime import datetime
+from fairseq import checkpoint_utils
+from lib.infer_pack.models import (
+    SynthesizerTrnMs256NSFsid,
+    SynthesizerTrnMs256NSFsid_nono,
+    SynthesizerTrnMs768NSFsid,
+    SynthesizerTrnMs768NSFsid_nono,
+)
+from vc_infer_pipeline import VC
+from config import Config
+config = Config()
+logging.getLogger("numba").setLevel(logging.WARNING)
+spaces = os.getenv("SYSTEM") == "spaces"
+force_support = None
+if config.unsupported is False:
+    if config.device == "mps" or config.device == "cpu":
+        force_support = False
+else:
+    force_support = True
+
+audio_mode = []
+f0method_mode = []
+f0method_info = ""
+
+if force_support is False or spaces is True:
+    if spaces is True:
+        audio_mode = ["Upload audio", "TTS Audio"]
+    else:
+        audio_mode = ["Input path", "Upload audio", "TTS Audio"]
+    f0method_mode = ["pm", "harvest"]
+    f0method_info = "PM is fast, Harvest is good but extremely slow, Rvmpe is alternative to harvest (might be better). (Default: PM)"
+else:
+    audio_mode = ["Input path", "Upload audio", "Youtube", "TTS Audio"]
+    f0method_mode = ["pm", "harvest", "crepe"]
+    f0method_info = "PM is fast, Harvest is good but extremely slow, Rvmpe is alternative to harvest (might be better), and Crepe effect is good but requires GPU (Default: PM)"
+
+if os.path.isfile("rmvpe.pt"):
+    f0method_mode.insert(2, "rmvpe")
+
+def create_vc_fn(model_name, tgt_sr, net_g, vc, if_f0, version, file_index):
+    def vc_fn(
+        vc_audio_mode,
+        vc_input, 
+        vc_upload,
+        tts_text,
+        tts_voice,
+        f0_up_key,
+        f0_method,
+        index_rate,
+        filter_radius,
+        resample_sr,
+        rms_mix_rate,
+        protect,
+    ):
+        try:
+            logs = []
+            print(f"Converting using {model_name}...")
+            logs.append(f"Converting using {model_name}...")
+            yield "\n".join(logs), None
+            if vc_audio_mode == "Input path" or "Youtube" and vc_input != "":
+                audio, sr = librosa.load(vc_input, sr=16000, mono=True)
+            elif vc_audio_mode == "Upload audio":
+                if vc_upload is None:
+                    return "You need to upload an audio", None
+                sampling_rate, audio = vc_upload
+                duration = audio.shape[0] / sampling_rate
+                if duration > 20 and spaces:
+                    return "Please upload an audio file that is less than 20 seconds. If you need to generate a longer audio file, please use Colab.", None
+                audio = (audio / np.iinfo(audio.dtype).max).astype(np.float32)
+                if len(audio.shape) > 1:
+                    audio = librosa.to_mono(audio.transpose(1, 0))
+                if sampling_rate != 16000:
+                    audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=16000)
+            elif vc_audio_mode == "TTS Audio":
+                if len(tts_text) > 100 and spaces:
+                    return "Text is too long", None
+                if tts_text is None or tts_voice is None:
+                    return "You need to enter text and select a voice", None
+                asyncio.run(edge_tts.Communicate(tts_text, "-".join(tts_voice.split('-')[:-1])).save("tts.mp3"))
+                audio, sr = librosa.load("tts.mp3", sr=16000, mono=True)
+                vc_input = "tts.mp3"
+            times = [0, 0, 0]
+            f0_up_key = int(f0_up_key)
+            audio_opt = vc.pipeline(
+                hubert_model,
+                net_g,
+                0,
+                audio,
+                vc_input,
+                times,
+                f0_up_key,
+                f0_method,
+                file_index,
+                # file_big_npy,
+                index_rate,
+                if_f0,
+                filter_radius,
+                tgt_sr,
+                resample_sr,
+                rms_mix_rate,
+                version,
+                protect,
+                f0_file=None,
+            )
+            info = f"[{datetime.now().strftime('%Y-%m-%d %H:%M')}]: npy: {times[0]}, f0: {times[1]}s, infer: {times[2]}s"
+            print(f"{model_name} | {info}")
+            logs.append(f"Successfully Convert {model_name}\n{info}")
+            yield "\n".join(logs), (tgt_sr, audio_opt)
+        except:
+            info = traceback.format_exc()
+            print(info)
+            yield info, None
+    return vc_fn
+
+def load_model():
+    categories = []
+    if os.path.isfile("weights/folder_info.json"):
+        with open("weights/folder_info.json", "r", encoding="utf-8") as f:
+            folder_info = json.load(f)
+        for category_name, category_info in folder_info.items():
+            if not category_info['enable']:
+                continue
+            category_title = category_info['title']
+            category_folder = category_info['folder_path']
+            description = category_info['description']
+            models = []
+            with open(f"weights/{category_folder}/model_info.json", "r", encoding="utf-8") as f:
+                models_info = json.load(f)
+            for character_name, info in models_info.items():
+                if not info['enable']:
+                    continue
+                model_title = info['title']
+                model_name = info['model_path']
+                model_author = info.get("author", None)
+                model_cover = f"weights/{category_folder}/{character_name}/{info['cover']}"
+                model_index = f"weights/{category_folder}/{character_name}/{info['feature_retrieval_library']}"
+                cpt = torch.load(f"weights/{category_folder}/{character_name}/{model_name}", 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)
+                version = cpt.get("version", "v1")
+                if version == "v1":
+                    if if_f0 == 1:
+                        net_g = SynthesizerTrnMs256NSFsid(*cpt["config"], is_half=config.is_half)
+                    else:
+                        net_g = SynthesizerTrnMs256NSFsid_nono(*cpt["config"])
+                    model_version = "V1"
+                elif version == "v2":
+                    if if_f0 == 1:
+                        net_g = SynthesizerTrnMs768NSFsid(*cpt["config"], is_half=config.is_half)
+                    else:
+                        net_g = SynthesizerTrnMs768NSFsid_nono(*cpt["config"])
+                    model_version = "V2"
+                del net_g.enc_q
+                print(net_g.load_state_dict(cpt["weight"], strict=False))
+                net_g.eval().to(config.device)
+                if config.is_half:
+                    net_g = net_g.half()
+                else:
+                    net_g = net_g.float()
+                vc = VC(tgt_sr, config)
+                print(f"Model loaded: {character_name} / {info['feature_retrieval_library']} | ({model_version})")
+                models.append((character_name, model_title, model_author, model_cover, model_version, create_vc_fn(model_name, tgt_sr, net_g, vc, if_f0, version, model_index)))
+            categories.append([category_title, category_folder, description, models])
+    else:
+        categories = []
+    return categories
+
+def download_audio(url, audio_provider):
+    logs = []
+    if url == "":
+        raise gr.Error("URL Required!")
+        return "URL Required"
+    if not os.path.exists("dl_audio"):
+        os.mkdir("dl_audio")
+    if audio_provider == "Youtube":
+        logs.append("Downloading the audio...")
+        yield None, "\n".join(logs)
+        ydl_opts = {
+            'noplaylist': True,
+            'format': 'bestaudio/best',
+            'postprocessors': [{
+                'key': 'FFmpegExtractAudio',
+                'preferredcodec': 'wav',
+            }],
+            "outtmpl": 'dl_audio/audio',
+        }
+        audio_path = "dl_audio/audio.wav"
+        with yt_dlp.YoutubeDL(ydl_opts) as ydl:
+            ydl.download([url])
+        logs.append("Download Complete.")
+        yield audio_path, "\n".join(logs)
+
+def cut_vocal_and_inst(split_model):
+    logs = []
+    logs.append("Starting the audio splitting process...")
+    yield "\n".join(logs), None, None, None, None
+    command = f"demucs --two-stems=vocals -n {split_model} dl_audio/audio.wav -o output"
+    result = subprocess.Popen(command.split(), stdout=subprocess.PIPE, text=True)
+    for line in result.stdout:
+        logs.append(line)
+        yield "\n".join(logs), None, None, None, None
+    print(result.stdout)
+    vocal = f"output/{split_model}/audio/vocals.wav"
+    inst = f"output/{split_model}/audio/no_vocals.wav"
+    logs.append("Audio splitting complete.")
+    yield "\n".join(logs), vocal, inst, vocal
+
+def combine_vocal_and_inst(audio_data, vocal_volume, inst_volume, split_model):
+    if not os.path.exists("output/result"):
+        os.mkdir("output/result")
+    vocal_path = "output/result/output.wav"
+    output_path = "output/result/combine.mp3"
+    inst_path = f"output/{split_model}/audio/no_vocals.wav"
+    with wave.open(vocal_path, "w") as wave_file:
+        wave_file.setnchannels(1) 
+        wave_file.setsampwidth(2)
+        wave_file.setframerate(audio_data[0])
+        wave_file.writeframes(audio_data[1].tobytes())
+    command =  f'ffmpeg -y -i {inst_path} -i {vocal_path} -filter_complex [0:a]volume={inst_volume}[i];[1:a]volume={vocal_volume}[v];[i][v]amix=inputs=2:duration=longest[a] -map [a] -b:a 320k -c:a libmp3lame {output_path}'
+    result = subprocess.run(command.split(), stdout=subprocess.PIPE)
+    print(result.stdout.decode())
+    return output_path
+
+def load_hubert():
+    global hubert_model
+    models, _, _ = checkpoint_utils.load_model_ensemble_and_task(
+        ["hubert_base.pt"],
+        suffix="",
+    )
+    hubert_model = models[0]
+    hubert_model = hubert_model.to(config.device)
+    if config.is_half:
+        hubert_model = hubert_model.half()
+    else:
+        hubert_model = hubert_model.float()
+    hubert_model.eval()
+
+def change_audio_mode(vc_audio_mode):
+    if vc_audio_mode == "Input path":
+        return (
+            # Input & Upload
+            gr.Textbox.update(visible=True),
+            gr.Checkbox.update(visible=False),
+            gr.Audio.update(visible=False),
+            # Youtube
+            gr.Dropdown.update(visible=False),
+            gr.Textbox.update(visible=False),
+            gr.Textbox.update(visible=False),
+            gr.Button.update(visible=False),
+            # Splitter
+            gr.Dropdown.update(visible=False),
+            gr.Textbox.update(visible=False),
+            gr.Button.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Slider.update(visible=False),
+            gr.Slider.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Button.update(visible=False),
+            # TTS
+            gr.Textbox.update(visible=False),
+            gr.Dropdown.update(visible=False)
+        )
+    elif vc_audio_mode == "Upload audio":
+        return (
+            # Input & Upload
+            gr.Textbox.update(visible=False),
+            gr.Checkbox.update(visible=True),
+            gr.Audio.update(visible=True),
+            # Youtube
+            gr.Dropdown.update(visible=False),
+            gr.Textbox.update(visible=False),
+            gr.Textbox.update(visible=False),
+            gr.Button.update(visible=False),
+            # Splitter
+            gr.Dropdown.update(visible=False),
+            gr.Textbox.update(visible=False),
+            gr.Button.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Slider.update(visible=False),
+            gr.Slider.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Button.update(visible=False),
+            # TTS
+            gr.Textbox.update(visible=False),
+            gr.Dropdown.update(visible=False)
+        )
+    elif vc_audio_mode == "Youtube":
+        return (
+            # Input & Upload
+            gr.Textbox.update(visible=False),
+            gr.Checkbox.update(visible=False),
+            gr.Audio.update(visible=False),
+            # Youtube
+            gr.Dropdown.update(visible=True),
+            gr.Textbox.update(visible=True),
+            gr.Textbox.update(visible=True),
+            gr.Button.update(visible=True),
+            # Splitter
+            gr.Dropdown.update(visible=True),
+            gr.Textbox.update(visible=True),
+            gr.Button.update(visible=True),
+            gr.Audio.update(visible=True),
+            gr.Audio.update(visible=True),
+            gr.Audio.update(visible=True),
+            gr.Slider.update(visible=True),
+            gr.Slider.update(visible=True),
+            gr.Audio.update(visible=True),
+            gr.Button.update(visible=True),
+            # TTS
+            gr.Textbox.update(visible=False),
+            gr.Dropdown.update(visible=False)
+        )
+    elif vc_audio_mode == "TTS Audio":
+        return (
+            # Input & Upload
+            gr.Textbox.update(visible=False),
+            gr.Checkbox.update(visible=False),
+            gr.Audio.update(visible=False),
+            # Youtube
+            gr.Dropdown.update(visible=False),
+            gr.Textbox.update(visible=False),
+            gr.Textbox.update(visible=False),
+            gr.Button.update(visible=False),
+            # Splitter
+            gr.Dropdown.update(visible=False),
+            gr.Textbox.update(visible=False),
+            gr.Button.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Slider.update(visible=False),
+            gr.Slider.update(visible=False),
+            gr.Audio.update(visible=False),
+            gr.Button.update(visible=False),
+            # TTS
+            gr.Textbox.update(visible=True),
+            gr.Dropdown.update(visible=True)
+        )
+
+def use_microphone(microphone):
+    if microphone == True:
+        return gr.Audio.update(source="microphone")
+    else:
+        return gr.Audio.update(source="upload")
+
+if __name__ == '__main__':
+    load_hubert()
+    categories = load_model()
+    tts_voice_list = asyncio.new_event_loop().run_until_complete(edge_tts.list_voices())
+    voices = [f"{v['ShortName']}-{v['Gender']}" for v in tts_voice_list]
+    with gr.Blocks() as app:
+        gr.Markdown(
+            "<div align='center'>\n\n"+
+            "# Multi Model RVC Inference\n\n"+
+            "[![Repository](https://img.shields.io/badge/Github-Multi%20Model%20RVC%20Inference-blue?style=for-the-badge&logo=github)](https://github.com/ArkanDash/Multi-Model-RVC-Inference)\n\n"+
+            "</div>"
+        )
+        if categories == []:
+            gr.Markdown(
+                "<div align='center'>\n\n"+
+                "## No model found, please add the model into weights folder\n\n"+
+                "</div>"
+            )
+        for (folder_title, folder, description, models) in categories:
+            with gr.TabItem(folder_title):
+                if description:
+                    gr.Markdown(f"### <center> {description}")
+                with gr.Tabs():
+                    if not models:
+                        gr.Markdown("# <center> No Model Loaded.")
+                        gr.Markdown("## <center> Please add the model or fix your model path.")
+                        continue
+                    for (name, title, author, cover, model_version, vc_fn) in models:
+                        with gr.TabItem(name):
+                            with gr.Row():
+                                gr.Markdown(
+                                    '<div align="center">'
+                                    f'<div>{title}</div>\n'+
+                                    f'<div>RVC {model_version} Model</div>\n'+
+                                    (f'<div>Model author: {author}</div>' if author else "")+
+                                    (f'<img style="width:auto;height:300px;" src="file/{cover}">' if cover else "")+
+                                    '</div>'
+                                )
+                            with gr.Row():
+                                if spaces is False:
+                                    with gr.TabItem("Input"):
+                                        with gr.Row():
+                                            with gr.Column():
+                                                vc_audio_mode = gr.Dropdown(label="Input voice", choices=audio_mode, allow_custom_value=False, value="Upload audio")
+                                                # Input
+                                                vc_input = gr.Textbox(label="Input audio path", visible=False)
+                                                # Upload
+                                                vc_microphone_mode = gr.Checkbox(label="Use Microphone", value=False, visible=True, interactive=True)
+                                                vc_upload = gr.Audio(label="Upload audio file", source="upload", visible=True, interactive=True)
+                                                # Youtube
+                                                vc_download_audio = gr.Dropdown(label="Provider", choices=["Youtube"], allow_custom_value=False, visible=False, value="Youtube", info="Select provider (Default: Youtube)")
+                                                vc_link = gr.Textbox(label="Youtube URL", visible=False, info="Example: https://www.youtube.com/watch?v=Nc0sB1Bmf-A", placeholder="https://www.youtube.com/watch?v=...")
+                                                vc_log_yt = gr.Textbox(label="Output Information", visible=False, interactive=False)
+                                                vc_download_button = gr.Button("Download Audio", variant="primary", visible=False)
+                                                vc_audio_preview = gr.Audio(label="Audio Preview", visible=False)
+                                                # TTS
+                                                tts_text = gr.Textbox(label="TTS text", info="Text to speech input", visible=False)
+                                                tts_voice = gr.Dropdown(label="Edge-tts speaker", choices=voices, visible=False, allow_custom_value=False, value="en-US-AnaNeural-Female")
+                                            with gr.Column():
+                                                vc_split_model = gr.Dropdown(label="Splitter Model", choices=["hdemucs_mmi", "htdemucs", "htdemucs_ft", "mdx", "mdx_q", "mdx_extra_q"], allow_custom_value=False, visible=False, value="htdemucs", info="Select the splitter model (Default: htdemucs)")
+                                                vc_split_log = gr.Textbox(label="Output Information", visible=False, interactive=False)
+                                                vc_split = gr.Button("Split Audio", variant="primary", visible=False)
+                                                vc_vocal_preview = gr.Audio(label="Vocal Preview", visible=False)
+                                                vc_inst_preview = gr.Audio(label="Instrumental Preview", visible=False)
+                                    with gr.TabItem("Convert"):
+                                        with gr.Row():
+                                            with gr.Column():
+                                                vc_transform0 = gr.Number(label="Transpose", value=0, info='Type "12" to change from male to female voice. Type "-12" to change female to male voice')
+                                                f0method0 = gr.Radio(
+                                                    label="Pitch extraction algorithm",
+                                                    info=f0method_info,
+                                                    choices=f0method_mode,
+                                                    value="pm",
+                                                    interactive=True
+                                                )
+                                                index_rate1 = gr.Slider(
+                                                    minimum=0,
+                                                    maximum=1,
+                                                    label="Retrieval feature ratio",
+                                                    info="(Default: 0.7)",
+                                                    value=0.7,
+                                                    interactive=True,
+                                                )
+                                                filter_radius0 = gr.Slider(
+                                                    minimum=0,
+                                                    maximum=7,
+                                                    label="Apply Median Filtering",
+                                                    info="The value represents the filter radius and can reduce breathiness.",
+                                                    value=3,
+                                                    step=1,
+                                                    interactive=True,
+                                                )
+                                                resample_sr0 = gr.Slider(
+                                                    minimum=0,
+                                                    maximum=48000,
+                                                    label="Resample the output audio",
+                                                    info="Resample the output audio in post-processing to the final sample rate. Set to 0 for no resampling",
+                                                    value=0,
+                                                    step=1,
+                                                    interactive=True,
+                                                )
+                                                rms_mix_rate0 = gr.Slider(
+                                                    minimum=0,
+                                                    maximum=1,
+                                                    label="Volume Envelope",
+                                                    info="Use the volume envelope of the input to replace or mix with the volume envelope of the output. The closer the ratio is to 1, the more the output envelope is used",
+                                                    value=1,
+                                                    interactive=True,
+                                                )
+                                                protect0 = gr.Slider(
+                                                    minimum=0,
+                                                    maximum=0.5,
+                                                    label="Voice Protection",
+                                                    info="Protect voiceless consonants and breath sounds to prevent artifacts such as tearing in electronic music. Set to 0.5 to disable. Decrease the value to increase protection, but it may reduce indexing accuracy",
+                                                    value=0.5,
+                                                    step=0.01,
+                                                    interactive=True,
+                                                )
+                                            with gr.Column():
+                                                vc_log = gr.Textbox(label="Output Information", interactive=False)
+                                                vc_output = gr.Audio(label="Output Audio", interactive=False)
+                                                vc_convert = gr.Button("Convert", variant="primary")
+                                                vc_vocal_volume = gr.Slider(
+                                                    minimum=0,
+                                                    maximum=10,
+                                                    label="Vocal volume",
+                                                    value=1,
+                                                    interactive=True,
+                                                    step=1,
+                                                    info="Adjust vocal volume (Default: 1}",
+                                                    visible=False
+                                                )
+                                                vc_inst_volume = gr.Slider(
+                                                    minimum=0,
+                                                    maximum=10,
+                                                    label="Instrument volume",
+                                                    value=1,
+                                                    interactive=True,
+                                                    step=1,
+                                                    info="Adjust instrument volume (Default: 1}",
+                                                    visible=False
+                                                )
+                                                vc_combined_output = gr.Audio(label="Output Combined Audio", visible=False)
+                                                vc_combine =  gr.Button("Combine",variant="primary", visible=False)
+                                else:
+                                    with gr.Column():
+                                        vc_audio_mode = gr.Dropdown(label="Input voice", choices=audio_mode, allow_custom_value=False, value="Upload audio")
+                                        # Input
+                                        vc_input = gr.Textbox(label="Input audio path", visible=False)
+                                        # Upload
+                                        vc_microphone_mode = gr.Checkbox(label="Use Microphone", value=False, visible=True, interactive=True)
+                                        vc_upload = gr.Audio(label="Upload audio file", source="upload", visible=True, interactive=True)
+                                        # Youtube
+                                        vc_download_audio = gr.Dropdown(label="Provider", choices=["Youtube"], allow_custom_value=False, visible=False, value="Youtube", info="Select provider (Default: Youtube)")
+                                        vc_link = gr.Textbox(label="Youtube URL", visible=False, info="Example: https://www.youtube.com/watch?v=Nc0sB1Bmf-A", placeholder="https://www.youtube.com/watch?v=...")
+                                        vc_log_yt = gr.Textbox(label="Output Information", visible=False, interactive=False)
+                                        vc_download_button = gr.Button("Download Audio", variant="primary", visible=False)
+                                        vc_audio_preview = gr.Audio(label="Audio Preview", visible=False)
+                                        # Splitter
+                                        vc_split_model = gr.Dropdown(label="Splitter Model", choices=["hdemucs_mmi", "htdemucs", "htdemucs_ft", "mdx", "mdx_q", "mdx_extra_q"], allow_custom_value=False, visible=False, value="htdemucs", info="Select the splitter model (Default: htdemucs)")
+                                        vc_split_log = gr.Textbox(label="Output Information", visible=False, interactive=False)
+                                        vc_split = gr.Button("Split Audio", variant="primary", visible=False)
+                                        vc_vocal_preview = gr.Audio(label="Vocal Preview", visible=False)
+                                        vc_inst_preview = gr.Audio(label="Instrumental Preview", visible=False)
+                                        # TTS
+                                        tts_text = gr.Textbox(label="TTS text", info="Text to speech input", visible=False)
+                                        tts_voice = gr.Dropdown(label="Edge-tts speaker", choices=voices, visible=False, allow_custom_value=False, value="en-US-AnaNeural-Female")
+                                    with gr.Column():
+                                        vc_transform0 = gr.Number(label="Transpose", value=0, info='Type "12" to change from male to female voice. Type "-12" to change female to male voice')
+                                        f0method0 = gr.Radio(
+                                            label="Pitch extraction algorithm",
+                                            info=f0method_info,
+                                            choices=f0method_mode,
+                                            value="pm",
+                                            interactive=True
+                                        )
+                                        index_rate1 = gr.Slider(
+                                            minimum=0,
+                                            maximum=1,
+                                            label="Retrieval feature ratio",
+                                            info="(Default: 0.7)",
+                                            value=0.7,
+                                            interactive=True,
+                                        )
+                                        filter_radius0 = gr.Slider(
+                                            minimum=0,
+                                            maximum=7,
+                                            label="Apply Median Filtering",
+                                            info="The value represents the filter radius and can reduce breathiness.",
+                                            value=3,
+                                            step=1,
+                                            interactive=True,
+                                        )
+                                        resample_sr0 = gr.Slider(
+                                            minimum=0,
+                                            maximum=48000,
+                                            label="Resample the output audio",
+                                            info="Resample the output audio in post-processing to the final sample rate. Set to 0 for no resampling",
+                                            value=0,
+                                            step=1,
+                                            interactive=True,
+                                        )
+                                        rms_mix_rate0 = gr.Slider(
+                                            minimum=0,
+                                            maximum=1,
+                                            label="Volume Envelope",
+                                            info="Use the volume envelope of the input to replace or mix with the volume envelope of the output. The closer the ratio is to 1, the more the output envelope is used",
+                                            value=1,
+                                            interactive=True,
+                                        )
+                                        protect0 = gr.Slider(
+                                            minimum=0,
+                                            maximum=0.5,
+                                            label="Voice Protection",
+                                            info="Protect voiceless consonants and breath sounds to prevent artifacts such as tearing in electronic music. Set to 0.5 to disable. Decrease the value to increase protection, but it may reduce indexing accuracy",
+                                            value=0.5,
+                                            step=0.01,
+                                            interactive=True,
+                                        )
+                                    with gr.Column():
+                                        vc_log = gr.Textbox(label="Output Information", interactive=False)
+                                        vc_output = gr.Audio(label="Output Audio", interactive=False)
+                                        vc_convert = gr.Button("Convert", variant="primary")
+                                        vc_vocal_volume = gr.Slider(
+                                            minimum=0,
+                                            maximum=10,
+                                            label="Vocal volume",
+                                            value=1,
+                                            interactive=True,
+                                            step=1,
+                                            info="Adjust vocal volume (Default: 1}",
+                                            visible=False
+                                        )
+                                        vc_inst_volume = gr.Slider(
+                                            minimum=0,
+                                            maximum=10,
+                                            label="Instrument volume",
+                                            value=1,
+                                            interactive=True,
+                                            step=1,
+                                            info="Adjust instrument volume (Default: 1}",
+                                            visible=False
+                                        )
+                                        vc_combined_output = gr.Audio(label="Output Combined Audio", visible=False)
+                                        vc_combine =  gr.Button("Combine",variant="primary", visible=False)
+                        vc_convert.click(
+                            fn=vc_fn, 
+                            inputs=[
+                                vc_audio_mode,
+                                vc_input,
+                                vc_upload,
+                                tts_text,
+                                tts_voice,
+                                vc_transform0,
+                                f0method0,
+                                index_rate1,
+                                filter_radius0,
+                                resample_sr0,
+                                rms_mix_rate0,
+                                protect0,
+                            ], 
+                            outputs=[vc_log ,vc_output]
+                        )
+                        vc_download_button.click(
+                            fn=download_audio, 
+                            inputs=[vc_link, vc_download_audio], 
+                            outputs=[vc_audio_preview, vc_log_yt]
+                        )
+                        vc_split.click(
+                            fn=cut_vocal_and_inst, 
+                            inputs=[vc_split_model], 
+                            outputs=[vc_split_log, vc_vocal_preview, vc_inst_preview, vc_input]
+                        )
+                        vc_combine.click(
+                            fn=combine_vocal_and_inst,
+                            inputs=[vc_output, vc_vocal_volume, vc_inst_volume, vc_split_model],
+                            outputs=[vc_combined_output]
+                        )
+                        vc_microphone_mode.change(
+                            fn=use_microphone,
+                            inputs=vc_microphone_mode,
+                            outputs=vc_upload
+                        )
+                        vc_audio_mode.change(
+                            fn=change_audio_mode,
+                            inputs=[vc_audio_mode],
+                            outputs=[
+                                vc_input,
+                                vc_microphone_mode,
+                                vc_upload,
+                                vc_download_audio,
+                                vc_link,
+                                vc_log_yt,
+                                vc_download_button,
+                                vc_split_model,
+                                vc_split_log,
+                                vc_split,
+                                vc_audio_preview,
+                                vc_vocal_preview,
+                                vc_inst_preview,
+                                vc_vocal_volume,
+                                vc_inst_volume,
+                                vc_combined_output,
+                                vc_combine,
+                                tts_text,
+                                tts_voice
+                            ]
+                        )
+        app.queue(concurrency_count=1, max_size=20, api_open=config.api).launch(share=config.colab)

config.py

@@ -1,4 +1,5 @@
 import argparse
+import sys
 import torch
 from multiprocessing import cpu_count
 
@@ -10,45 +11,38 @@ class Config:
         self.gpu_name = None
         self.gpu_mem = None
         (
-            self.python_cmd,
-            self.listen_port,
             self.colab,
-            self.noparallel,
-            self.noautoopen,
-            self.api
+            self.api,
+            self.unsupported
         ) = self.arg_parse()
         self.x_pad, self.x_query, self.x_center, self.x_max = self.device_config()
 
     @staticmethod
     def arg_parse() -> tuple:
         parser = argparse.ArgumentParser()
-        parser.add_argument("--port", type=int, default=7865, help="Listen port")
-        parser.add_argument(
-            "--pycmd", type=str, default="python", help="Python command"
-        )
         parser.add_argument("--colab", action="store_true", help="Launch in colab")
-        parser.add_argument(
-            "--noparallel", action="store_true", help="Disable parallel processing"
-        )
-        parser.add_argument(
-            "--noautoopen",
-            action="store_true",
-            help="Do not open in browser automatically",
-        )
         parser.add_argument("--api", action="store_true", help="Launch with api")
+        parser.add_argument("--unsupported", action="store_true", help="Enable unsupported feature")
         cmd_opts = parser.parse_args()
 
-        cmd_opts.port = cmd_opts.port if 0 <= cmd_opts.port <= 65535 else 7865
-
         return (
-            cmd_opts.pycmd,
-            cmd_opts.port,
             cmd_opts.colab,
-            cmd_opts.noparallel,
-            cmd_opts.noautoopen,
-            cmd_opts.api
+            cmd_opts.api,
+            cmd_opts.unsupported
         )
 
+    # has_mps is only available in nightly pytorch (for now) and MasOS 12.3+.
+    # check `getattr` and try it for compatibility
+    @staticmethod
+    def has_mps() -> bool:
+        if not torch.backends.mps.is_available():
+            return False
+        try:
+            torch.zeros(1).to(torch.device("mps"))
+            return True
+        except Exception:
+            return False
+
     def device_config(self) -> tuple:
         if torch.cuda.is_available():
             i_device = int(self.device.split(":")[-1])
@@ -60,11 +54,10 @@ class Config:
                 or "1070" in self.gpu_name
                 or "1080" in self.gpu_name
             ):
-                print("16系/10系显卡和P40强制单精度")
+                print("INFO: Found GPU", self.gpu_name, ", force to fp32")
                 self.is_half = False
-
             else:
-                self.gpu_name = None
+                print("INFO: Found GPU", self.gpu_name)
             self.gpu_mem = int(
                 torch.cuda.get_device_properties(i_device).total_memory
                 / 1024
@@ -72,12 +65,12 @@ class Config:
                 / 1024
                 + 0.4
             )
-        elif torch.backends.mps.is_available():
-            print("没有发现支持的N卡, 使用MPS进行推理")
+        elif self.has_mps():
+            print("INFO: No supported Nvidia GPU found, use MPS instead")
             self.device = "mps"
             self.is_half = False
         else:
-            print("没有发现支持的N卡, 使用CPU进行推理")
+            print("INFO: No supported Nvidia GPU found, use CPU instead")
             self.device = "cpu"
             self.is_half = False
 

rmvpe.py

@@ -0,0 +1,432 @@
+import sys, torch, numpy as np, traceback, pdb
+import torch.nn as nn
+from time import time as ttime
+import torch.nn.functional as F
+
+
+class BiGRU(nn.Module):
+    def __init__(self, input_features, hidden_features, num_layers):
+        super(BiGRU, self).__init__()
+        self.gru = nn.GRU(
+            input_features,
+            hidden_features,
+            num_layers=num_layers,
+            batch_first=True,
+            bidirectional=True,
+        )
+
+    def forward(self, x):
+        return self.gru(x)[0]
+
+
+class ConvBlockRes(nn.Module):
+    def __init__(self, in_channels, out_channels, momentum=0.01):
+        super(ConvBlockRes, self).__init__()
+        self.conv = nn.Sequential(
+            nn.Conv2d(
+                in_channels=in_channels,
+                out_channels=out_channels,
+                kernel_size=(3, 3),
+                stride=(1, 1),
+                padding=(1, 1),
+                bias=False,
+            ),
+            nn.BatchNorm2d(out_channels, momentum=momentum),
+            nn.ReLU(),
+            nn.Conv2d(
+                in_channels=out_channels,
+                out_channels=out_channels,
+                kernel_size=(3, 3),
+                stride=(1, 1),
+                padding=(1, 1),
+                bias=False,
+            ),
+            nn.BatchNorm2d(out_channels, momentum=momentum),
+            nn.ReLU(),
+        )
+        if in_channels != out_channels:
+            self.shortcut = nn.Conv2d(in_channels, out_channels, (1, 1))
+            self.is_shortcut = True
+        else:
+            self.is_shortcut = False
+
+    def forward(self, x):
+        if self.is_shortcut:
+            return self.conv(x) + self.shortcut(x)
+        else:
+            return self.conv(x) + x
+
+
+class Encoder(nn.Module):
+    def __init__(
+        self,
+        in_channels,
+        in_size,
+        n_encoders,
+        kernel_size,
+        n_blocks,
+        out_channels=16,
+        momentum=0.01,
+    ):
+        super(Encoder, self).__init__()
+        self.n_encoders = n_encoders
+        self.bn = nn.BatchNorm2d(in_channels, momentum=momentum)
+        self.layers = nn.ModuleList()
+        self.latent_channels = []
+        for i in range(self.n_encoders):
+            self.layers.append(
+                ResEncoderBlock(
+                    in_channels, out_channels, kernel_size, n_blocks, momentum=momentum
+                )
+            )
+            self.latent_channels.append([out_channels, in_size])
+            in_channels = out_channels
+            out_channels *= 2
+            in_size //= 2
+        self.out_size = in_size
+        self.out_channel = out_channels
+
+    def forward(self, x):
+        concat_tensors = []
+        x = self.bn(x)
+        for i in range(self.n_encoders):
+            _, x = self.layers[i](x)
+            concat_tensors.append(_)
+        return x, concat_tensors
+
+
+class ResEncoderBlock(nn.Module):
+    def __init__(
+        self, in_channels, out_channels, kernel_size, n_blocks=1, momentum=0.01
+    ):
+        super(ResEncoderBlock, self).__init__()
+        self.n_blocks = n_blocks
+        self.conv = nn.ModuleList()
+        self.conv.append(ConvBlockRes(in_channels, out_channels, momentum))
+        for i in range(n_blocks - 1):
+            self.conv.append(ConvBlockRes(out_channels, out_channels, momentum))
+        self.kernel_size = kernel_size
+        if self.kernel_size is not None:
+            self.pool = nn.AvgPool2d(kernel_size=kernel_size)
+
+    def forward(self, x):
+        for i in range(self.n_blocks):
+            x = self.conv[i](x)
+        if self.kernel_size is not None:
+            return x, self.pool(x)
+        else:
+            return x
+
+
+class Intermediate(nn.Module):  #
+    def __init__(self, in_channels, out_channels, n_inters, n_blocks, momentum=0.01):
+        super(Intermediate, self).__init__()
+        self.n_inters = n_inters
+        self.layers = nn.ModuleList()
+        self.layers.append(
+            ResEncoderBlock(in_channels, out_channels, None, n_blocks, momentum)
+        )
+        for i in range(self.n_inters - 1):
+            self.layers.append(
+                ResEncoderBlock(out_channels, out_channels, None, n_blocks, momentum)
+            )
+
+    def forward(self, x):
+        for i in range(self.n_inters):
+            x = self.layers[i](x)
+        return x
+
+
+class ResDecoderBlock(nn.Module):
+    def __init__(self, in_channels, out_channels, stride, n_blocks=1, momentum=0.01):
+        super(ResDecoderBlock, self).__init__()
+        out_padding = (0, 1) if stride == (1, 2) else (1, 1)
+        self.n_blocks = n_blocks
+        self.conv1 = nn.Sequential(
+            nn.ConvTranspose2d(
+                in_channels=in_channels,
+                out_channels=out_channels,
+                kernel_size=(3, 3),
+                stride=stride,
+                padding=(1, 1),
+                output_padding=out_padding,
+                bias=False,
+            ),
+            nn.BatchNorm2d(out_channels, momentum=momentum),
+            nn.ReLU(),
+        )
+        self.conv2 = nn.ModuleList()
+        self.conv2.append(ConvBlockRes(out_channels * 2, out_channels, momentum))
+        for i in range(n_blocks - 1):
+            self.conv2.append(ConvBlockRes(out_channels, out_channels, momentum))
+
+    def forward(self, x, concat_tensor):
+        x = self.conv1(x)
+        x = torch.cat((x, concat_tensor), dim=1)
+        for i in range(self.n_blocks):
+            x = self.conv2[i](x)
+        return x
+
+
+class Decoder(nn.Module):
+    def __init__(self, in_channels, n_decoders, stride, n_blocks, momentum=0.01):
+        super(Decoder, self).__init__()
+        self.layers = nn.ModuleList()
+        self.n_decoders = n_decoders
+        for i in range(self.n_decoders):
+            out_channels = in_channels // 2
+            self.layers.append(
+                ResDecoderBlock(in_channels, out_channels, stride, n_blocks, momentum)
+            )
+            in_channels = out_channels
+
+    def forward(self, x, concat_tensors):
+        for i in range(self.n_decoders):
+            x = self.layers[i](x, concat_tensors[-1 - i])
+        return x
+
+
+class DeepUnet(nn.Module):
+    def __init__(
+        self,
+        kernel_size,
+        n_blocks,
+        en_de_layers=5,
+        inter_layers=4,
+        in_channels=1,
+        en_out_channels=16,
+    ):
+        super(DeepUnet, self).__init__()
+        self.encoder = Encoder(
+            in_channels, 128, en_de_layers, kernel_size, n_blocks, en_out_channels
+        )
+        self.intermediate = Intermediate(
+            self.encoder.out_channel // 2,
+            self.encoder.out_channel,
+            inter_layers,
+            n_blocks,
+        )
+        self.decoder = Decoder(
+            self.encoder.out_channel, en_de_layers, kernel_size, n_blocks
+        )
+
+    def forward(self, x):
+        x, concat_tensors = self.encoder(x)
+        x = self.intermediate(x)
+        x = self.decoder(x, concat_tensors)
+        return x
+
+
+class E2E(nn.Module):
+    def __init__(
+        self,
+        n_blocks,
+        n_gru,
+        kernel_size,
+        en_de_layers=5,
+        inter_layers=4,
+        in_channels=1,
+        en_out_channels=16,
+    ):
+        super(E2E, self).__init__()
+        self.unet = DeepUnet(
+            kernel_size,
+            n_blocks,
+            en_de_layers,
+            inter_layers,
+            in_channels,
+            en_out_channels,
+        )
+        self.cnn = nn.Conv2d(en_out_channels, 3, (3, 3), padding=(1, 1))
+        if n_gru:
+            self.fc = nn.Sequential(
+                BiGRU(3 * 128, 256, n_gru),
+                nn.Linear(512, 360),
+                nn.Dropout(0.25),
+                nn.Sigmoid(),
+            )
+        else:
+            self.fc = nn.Sequential(
+                nn.Linear(3 * N_MELS, N_CLASS), nn.Dropout(0.25), nn.Sigmoid()
+            )
+
+    def forward(self, mel):
+        mel = mel.transpose(-1, -2).unsqueeze(1)
+        x = self.cnn(self.unet(mel)).transpose(1, 2).flatten(-2)
+        x = self.fc(x)
+        return x
+
+
+from librosa.filters import mel
+
+
+class MelSpectrogram(torch.nn.Module):
+    def __init__(
+        self,
+        is_half,
+        n_mel_channels,
+        sampling_rate,
+        win_length,
+        hop_length,
+        n_fft=None,
+        mel_fmin=0,
+        mel_fmax=None,
+        clamp=1e-5,
+    ):
+        super().__init__()
+        n_fft = win_length if n_fft is None else n_fft
+        self.hann_window = {}
+        mel_basis = mel(
+            sr=sampling_rate,
+            n_fft=n_fft,
+            n_mels=n_mel_channels,
+            fmin=mel_fmin,
+            fmax=mel_fmax,
+            htk=True,
+        )
+        mel_basis = torch.from_numpy(mel_basis).float()
+        self.register_buffer("mel_basis", mel_basis)
+        self.n_fft = win_length if n_fft is None else n_fft
+        self.hop_length = hop_length
+        self.win_length = win_length
+        self.sampling_rate = sampling_rate
+        self.n_mel_channels = n_mel_channels
+        self.clamp = clamp
+        self.is_half = is_half
+
+    def forward(self, audio, keyshift=0, speed=1, center=True):
+        factor = 2 ** (keyshift / 12)
+        n_fft_new = int(np.round(self.n_fft * factor))
+        win_length_new = int(np.round(self.win_length * factor))
+        hop_length_new = int(np.round(self.hop_length * speed))
+        keyshift_key = str(keyshift) + "_" + str(audio.device)
+        if keyshift_key not in self.hann_window:
+            self.hann_window[keyshift_key] = torch.hann_window(win_length_new).to(
+                audio.device
+            )
+        fft = torch.stft(
+            audio,
+            n_fft=n_fft_new,
+            hop_length=hop_length_new,
+            win_length=win_length_new,
+            window=self.hann_window[keyshift_key],
+            center=center,
+            return_complex=True,
+        )
+        magnitude = torch.sqrt(fft.real.pow(2) + fft.imag.pow(2))
+        if keyshift != 0:
+            size = self.n_fft // 2 + 1
+            resize = magnitude.size(1)
+            if resize < size:
+                magnitude = F.pad(magnitude, (0, 0, 0, size - resize))
+            magnitude = magnitude[:, :size, :] * self.win_length / win_length_new
+        mel_output = torch.matmul(self.mel_basis, magnitude)
+        if self.is_half == True:
+            mel_output = mel_output.half()
+        log_mel_spec = torch.log(torch.clamp(mel_output, min=self.clamp))
+        return log_mel_spec
+
+
+class RMVPE:
+    def __init__(self, model_path, is_half, device=None):
+        self.resample_kernel = {}
+        model = E2E(4, 1, (2, 2))
+        ckpt = torch.load(model_path, map_location="cpu")
+        model.load_state_dict(ckpt)
+        model.eval()
+        if is_half == True:
+            model = model.half()
+        self.model = model
+        self.resample_kernel = {}
+        self.is_half = is_half
+        if device is None:
+            device = "cuda" if torch.cuda.is_available() else "cpu"
+        self.device = device
+        self.mel_extractor = MelSpectrogram(
+            is_half, 128, 16000, 1024, 160, None, 30, 8000
+        ).to(device)
+        self.model = self.model.to(device)
+        cents_mapping = 20 * np.arange(360) + 1997.3794084376191
+        self.cents_mapping = np.pad(cents_mapping, (4, 4))  # 368
+
+    def mel2hidden(self, mel):
+        with torch.no_grad():
+            n_frames = mel.shape[-1]
+            mel = F.pad(
+                mel, (0, 32 * ((n_frames - 1) // 32 + 1) - n_frames), mode="reflect"
+            )
+            hidden = self.model(mel)
+            return hidden[:, :n_frames]
+
+    def decode(self, hidden, thred=0.03):
+        cents_pred = self.to_local_average_cents(hidden, thred=thred)
+        f0 = 10 * (2 ** (cents_pred / 1200))
+        f0[f0 == 10] = 0
+        # f0 = np.array([10 * (2 ** (cent_pred / 1200)) if cent_pred else 0 for cent_pred in cents_pred])
+        return f0
+
+    def infer_from_audio(self, audio, thred=0.03):
+        audio = torch.from_numpy(audio).float().to(self.device).unsqueeze(0)
+        # torch.cuda.synchronize()
+        # t0=ttime()
+        mel = self.mel_extractor(audio, center=True)
+        # torch.cuda.synchronize()
+        # t1=ttime()
+        hidden = self.mel2hidden(mel)
+        # torch.cuda.synchronize()
+        # t2=ttime()
+        hidden = hidden.squeeze(0).cpu().numpy()
+        if self.is_half == True:
+            hidden = hidden.astype("float32")
+        f0 = self.decode(hidden, thred=thred)
+        # torch.cuda.synchronize()
+        # t3=ttime()
+        # print("hmvpe:%s\t%s\t%s\t%s"%(t1-t0,t2-t1,t3-t2,t3-t0))
+        return f0
+
+    def to_local_average_cents(self, salience, thred=0.05):
+        # t0 = ttime()
+        center = np.argmax(salience, axis=1)  # 帧长#index
+        salience = np.pad(salience, ((0, 0), (4, 4)))  # 帧长,368
+        # t1 = ttime()
+        center += 4
+        todo_salience = []
+        todo_cents_mapping = []
+        starts = center - 4
+        ends = center + 5
+        for idx in range(salience.shape[0]):
+            todo_salience.append(salience[:, starts[idx] : ends[idx]][idx])
+            todo_cents_mapping.append(self.cents_mapping[starts[idx] : ends[idx]])
+        # t2 = ttime()
+        todo_salience = np.array(todo_salience)  # 帧长，9
+        todo_cents_mapping = np.array(todo_cents_mapping)  # 帧长，9
+        product_sum = np.sum(todo_salience * todo_cents_mapping, 1)
+        weight_sum = np.sum(todo_salience, 1)  # 帧长
+        devided = product_sum / weight_sum  # 帧长
+        # t3 = ttime()
+        maxx = np.max(salience, axis=1)  # 帧长
+        devided[maxx <= thred] = 0
+        # t4 = ttime()
+        # print("decode:%s\t%s\t%s\t%s" % (t1 - t0, t2 - t1, t3 - t2, t4 - t3))
+        return devided
+
+
+# if __name__ == '__main__':
+#     audio, sampling_rate = sf.read("卢本伟语录~1.wav")
+#     if len(audio.shape) > 1:
+#         audio = librosa.to_mono(audio.transpose(1, 0))
+#     audio_bak = audio.copy()
+#     if sampling_rate != 16000:
+#         audio = librosa.resample(audio, orig_sr=sampling_rate, target_sr=16000)
+#     model_path = "/bili-coeus/jupyter/jupyterhub-liujing04/vits_ch/test-RMVPE/weights/rmvpe_llc_half.pt"
+#     thred = 0.03  # 0.01
+#     device = 'cuda' if torch.cuda.is_available() else 'cpu'
+#     rmvpe = RMVPE(model_path,is_half=False, device=device)
+#     t0=ttime()
+#     f0 = rmvpe.infer_from_audio(audio, thred=thred)
+#     f0 = rmvpe.infer_from_audio(audio, thred=thred)
+#     f0 = rmvpe.infer_from_audio(audio, thred=thred)
+#     f0 = rmvpe.infer_from_audio(audio, thred=thred)
+#     f0 = rmvpe.infer_from_audio(audio, thred=thred)
+#     t1=ttime()
+#     print(f0.shape,t1-t0)

vc_infer_pipeline.py

@@ -1,4 +1,4 @@
-import numpy as np, parselmouth, torch, pdb
+import numpy as np, parselmouth, torch, pdb, sys, os
 from time import time as ttime
 import torch.nn.functional as F
 import scipy.signal as signal
@@ -6,6 +6,9 @@ import pyworld, os, traceback, faiss, librosa, torchcrepe
 from scipy import signal
 from functools import lru_cache
 
+now_dir = os.getcwd()
+sys.path.append(now_dir)
+
 bh, ah = signal.butter(N=5, Wn=48, btype="high", fs=16000)
 
 input_audio_path2wav = {}
@@ -124,6 +127,15 @@ class VC(object):
             f0 = torchcrepe.filter.mean(f0, 3)
             f0[pd < 0.1] = 0
             f0 = f0[0].cpu().numpy()
+        elif f0_method == "rmvpe":
+            if hasattr(self, "model_rmvpe") == False:
+                from rmvpe import RMVPE
+
+                print("loading rmvpe model")
+                self.model_rmvpe = RMVPE(
+                    "rmvpe.pt", is_half=self.is_half, device=self.device
+                )
+            f0 = self.model_rmvpe.infer_from_audio(x, thred=0.03)
         f0 *= pow(2, f0_up_key / 12)
         # with open("test.txt","w")as f:f.write("\n".join([str(i)for i in f0.tolist()]))
         tf0 = self.sr // self.window  # 每秒f0点数