Spaces:
No application file
No application file
File size: 5,517 Bytes
b5dba8a |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 |
{
"cells": [
{
"cell_type": "markdown",
"metadata": {
"collapsed": false
},
"source": [
"# Bark text-to-speech voice cloning.\n",
"Clone voices to create speaker history prompt files (.npz) for [bark text-to-speech](https://github.com/suno-ai/bark).\n",
"(This version of the notebook is made to work on Google Colab, make sure your runtime hardware accelerator is set to GPU)"
]
},
{
"cell_type": "markdown",
"metadata": {
"collapsed": false
},
"source": [
"# Google Colab: Clone the repository"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": false
},
"outputs": [],
"source": [
"!git clone https://github.com/gitmylo/bark-voice-cloning-HuBERT-quantizer/\n",
"%cd bark-voice-cloning-HuBERT-quantizer"
]
},
{
"cell_type": "markdown",
"metadata": {
"collapsed": false
},
"source": [
"## Install packages"
]
},
{
"cell_type": "code",
"execution_count": null,
"metadata": {
"collapsed": true
},
"outputs": [],
"source": [
"%pip install -r requirements.txt\n",
"%pip install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/cu117"
]
},
{
"cell_type": "markdown",
"metadata": {
"collapsed": false
},
"source": [
"## Load models"
]
},
{
"cell_type": "code",
"execution_count": 1,
"metadata": {
"collapsed": false
},
"outputs": [
{
"name": "stderr",
"output_type": "stream",
"text": [
"2023-05-26 21:27:49 | INFO | fairseq.tasks.text_to_speech | Please install tensorboardX: pip install tensorboardX\n"
]
},
{
"name": "stdout",
"output_type": "stream",
"text": [
"Loading HuBERT...\n",
"Loading Quantizer...\n",
"Loading Encodec...\n",
"Downloaded and loaded models!\n"
]
}
],
"source": [
"large_quant_model = False # Use the larger pretrained model\n",
"device = 'cuda' # 'cuda', 'cpu', 'cuda:0', 0, -1, torch.device('cuda')\n",
"\n",
"import numpy as np\n",
"import torch\n",
"import torchaudio\n",
"from encodec import EncodecModel\n",
"from encodec.utils import convert_audio\n",
"from bark_hubert_quantizer.hubert_manager import HuBERTManager\n",
"from bark_hubert_quantizer.pre_kmeans_hubert import CustomHubert\n",
"from bark_hubert_quantizer.customtokenizer import CustomTokenizer\n",
"\n",
"model = ('quantifier_V1_hubert_base_ls960_23.pth', 'tokenizer_large.pth') if large_quant_model else ('quantifier_hubert_base_ls960_14.pth', 'tokenizer.pth')\n",
"\n",
"print('Loading HuBERT...')\n",
"hubert_model = CustomHubert(HuBERTManager.make_sure_hubert_installed(), device=device)\n",
"print('Loading Quantizer...')\n",
"quant_model = CustomTokenizer.load_from_checkpoint(HuBERTManager.make_sure_tokenizer_installed(model=model[0], local_file=model[1]), device)\n",
"print('Loading Encodec...')\n",
"encodec_model = EncodecModel.encodec_model_24khz()\n",
"encodec_model.set_target_bandwidth(6.0)\n",
"encodec_model.to(device)\n",
"\n",
"print('Downloaded and loaded models!')"
]
},
{
"cell_type": "markdown",
"metadata": {
"collapsed": false
},
"source": [
"## Load wav and create speaker history prompt"
]
},
{
"cell_type": "code",
"execution_count": 2,
"metadata": {
"collapsed": false
},
"outputs": [
{
"name": "stdout",
"output_type": "stream",
"text": [
"Extracting semantics...\n",
"Tokenizing semantics...\n",
"Creating coarse and fine prompts...\n",
"Done!\n"
]
}
],
"source": [
"wav_file = 'speaker.wav' # Put the path of the speaker you want to use here.\n",
"out_file = 'speaker.npz' # Put the path to save the cloned speaker to here.\n",
"\n",
"wav, sr = torchaudio.load(wav_file)\n",
"\n",
"wav_hubert = wav.to(device)\n",
"\n",
"if wav_hubert.shape[0] == 2: # Stereo to mono if needed\n",
" wav_hubert = wav_hubert.mean(0, keepdim=True)\n",
"\n",
"print('Extracting semantics...')\n",
"semantic_vectors = hubert_model.forward(wav_hubert, input_sample_hz=sr)\n",
"print('Tokenizing semantics...')\n",
"semantic_tokens = quant_model.get_token(semantic_vectors)\n",
"print('Creating coarse and fine prompts...')\n",
"wav = convert_audio(wav, sr, encodec_model.sample_rate, 1).unsqueeze(0)\n",
"\n",
"wav = wav.to(device)\n",
"\n",
"with torch.no_grad():\n",
" encoded_frames = encodec_model.encode(wav)\n",
"codes = torch.cat([encoded[0] for encoded in encoded_frames], dim=-1).squeeze()\n",
"\n",
"codes = codes.cpu()\n",
"semantic_tokens = semantic_tokens.cpu()\n",
"\n",
"np.savez(out_file,\n",
" semantic_prompt=semantic_tokens,\n",
" fine_prompt=codes,\n",
" coarse_prompt=codes[:2, :]\n",
" )\n",
"\n",
"print('Done!')"
]
}
],
"metadata": {
"kernelspec": {
"display_name": "Python 3",
"language": "python",
"name": "python3"
},
"language_info": {
"codemirror_mode": {
"name": "ipython",
"version": 2
},
"file_extension": ".py",
"mimetype": "text/x-python",
"name": "python",
"nbconvert_exporter": "python",
"pygments_lexer": "ipython2",
"version": "2.7.6"
}
},
"nbformat": 4,
"nbformat_minor": 0
}
|