Paolify/RVC_4

1import gc2import requests3import subprocess4import logging5import sys6from bs4 import BeautifulSoup7import torch, pdb, os, warnings, librosa8import soundfile as sf9from tqdm import tqdm10import numpy as np11import torch12now_dir = os.getcwd()13sys.path.append(now_dir)14import mdx15branch = "https://github.com/NaJeongMo/Colab-for-MDX_B"16 17model_params = "https://raw.githubusercontent.com/TRvlvr/application_data/main/mdx_model_data/model_data.json"18_Models = "https://github.com/TRvlvr/model_repo/releases/download/all_public_uvr_models/"19# _models = "https://pastebin.com/raw/jBzYB8vz"20_models = "https://raw.githubusercontent.com/TRvlvr/application_data/main/filelists/download_checks.json"21#stem_naming = "https://pastebin.com/raw/mpH4hRcF"22 23file_folder = "Colab-for-MDX_B"24model_ids = requests.get(_models).json()25model_ids = model_ids["mdx_download_list"].values()26#print(model_ids)27model_params = requests.get(model_params).json()28#stem_naming = requests.get(stem_naming).json()29stem_naming = {30    "Vocals": "Instrumental",31    "Other": "Instruments",32    "Instrumental": "Vocals",33    "Drums": "Drumless",34    "Bass": "Bassless"35}36 37os.makedirs("tmp_models", exist_ok=True)38 39warnings.filterwarnings("ignore")40cpu = torch.device("cpu")41if torch.cuda.is_available():42    device = torch.device("cuda:0")43elif torch.backends.mps.is_available():44    device = torch.device("mps")45else:46    device = torch.device("cpu")47 48 49def get_model_list():50    return model_ids51 52def id_to_ptm(mkey):53    if mkey in model_ids:54        mpath = f"{now_dir}/tmp_models/{mkey}"55        if not os.path.exists(f'{now_dir}/tmp_models/{mkey}'):56            print('Downloading model...',end=' ')57            subprocess.run(58                ["wget", _Models+mkey, "-O", mpath]59            )60            print(f'saved to {mpath}')61            # get_ipython().system(f'gdown {model_id} -O /content/tmp_models/{mkey}')62            return mpath63        else:64            return mpath65    else:66        mpath = f'models/{mkey}'67        return mpath68 69def prepare_mdx(onnx,custom_param=False, dim_f=None, dim_t=None, n_fft=None, stem_name=None, compensation=None):70    device = torch.device('cuda:0') if torch.cuda.is_available() else torch.device('cpu')71    if custom_param:72        assert not (dim_f is None or dim_t is None or n_fft is None or compensation is None), 'Custom parameter selected, but incomplete parameters are provided.'73        mdx_model = mdx.MDX_Model(74            device,75            dim_f = dim_f,76            dim_t = dim_t,77            n_fft = n_fft,78            stem_name=stem_name,79            compensation=compensation80        )81    else:82        model_hash = mdx.MDX.get_hash(onnx)83        if model_hash in model_params:84            mp = model_params.get(model_hash)85            mdx_model = mdx.MDX_Model(86                device,87                dim_f = mp["mdx_dim_f_set"],88                dim_t = 2**mp["mdx_dim_t_set"],89                n_fft = mp["mdx_n_fft_scale_set"],90                stem_name=mp["primary_stem"],91                compensation=compensation if not custom_param and compensation is not None else mp["compensate"]92            )93    return mdx_model94 95def run_mdx(onnx, mdx_model,filename, output_format='wav',diff=False,suffix=None,diff_suffix=None, denoise=False, m_threads=2):96    mdx_sess = mdx.MDX(onnx,mdx_model)97    print(f"Processing: {filename}")98    if filename.lower().endswith('.wav'):99        wave, sr = librosa.load(filename, mono=False, sr=44100)100    else:101        temp_wav = 'temp_audio.wav'102        subprocess.run(['ffmpeg', '-i', filename, '-ar', '44100', '-ac', '2', temp_wav])  # Convert to WAV format103        wave, sr = librosa.load(temp_wav, mono=False, sr=44100)104        os.remove(temp_wav)105    106    #wave, sr = librosa.load(filename,mono=False, sr=44100)107    # normalizing input wave gives better output108    peak = max(np.max(wave), abs(np.min(wave)))109    wave /= peak110    if denoise:111        wave_processed = -(mdx_sess.process_wave(-wave, m_threads)) + (mdx_sess.process_wave(wave, m_threads))112        wave_processed *= 0.5113    else:114        wave_processed = mdx_sess.process_wave(wave, m_threads)115    # return to previous peak116    wave_processed *= peak117 118    stem_name = mdx_model.stem_name if suffix is None else suffix # use suffix if provided119    save_path = os.path.basename(os.path.splitext(filename)[0])120    #vocals_save_path = os.path.join(vocals_folder, f"{save_path}_{stem_name}.{output_format}")121    #instrumental_save_path = os.path.join(instrumental_folder, f"{save_path}_{stem_name}.{output_format}")122    save_path = f"{os.path.basename(os.path.splitext(filename)[0])}_{stem_name}.{output_format}"123    save_path = os.path.join(124            'audios',125            save_path126        )127    sf.write(128        save_path,129        wave_processed.T,130        sr131    )132 133    print(f'done, saved to: {save_path}')134 135    if diff:136        diff_stem_name = stem_naming.get(stem_name) if diff_suffix is None else diff_suffix # use suffix if provided137        stem_name = f"{stem_name}_diff" if diff_stem_name is None else diff_stem_name138        save_path = f"{os.path.basename(os.path.splitext(filename)[0])}_{stem_name}.{output_format}"139        save_path = os.path.join(140                'audio-others',141                save_path142            )143        sf.write(144            save_path,145            (-wave_processed.T*mdx_model.compensation)+wave.T,146            sr147        )148        print(f'invert done, saved to: {save_path}')149    del mdx_sess, wave_processed, wave150    gc.collect()151 152if __name__ == "__main__":153    print()