File size: 6,103 Bytes
7bcf8d7
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
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
{
 "cells": [
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "%matplotlib inline\n",
    "import matplotlib.pyplot as plt\n",
    "import IPython.display as ipd\n",
    "\n",
    "import os\n",
    "import json\n",
    "import math\n",
    "import torch\n",
    "from torch import nn\n",
    "from torch.nn import functional as F\n",
    "from torch.utils.data import DataLoader\n",
    "\n",
    "import commons\n",
    "import utils\n",
    "from data_utils import TextAudioLoader, TextAudioCollate, TextAudioSpeakerLoader, TextAudioSpeakerCollate\n",
    "from models import SynthesizerTrn\n",
    "from text.symbols import symbols\n",
    "from text import text_to_sequence\n",
    "\n",
    "from scipy.io.wavfile import write\n",
    "\n",
    "\n",
    "def get_text(text, hps):\n",
    "    text_norm = text_to_sequence(text, hps.data.text_cleaners)\n",
    "    if hps.data.add_blank:\n",
    "        text_norm = commons.intersperse(text_norm, 0)\n",
    "    text_norm = torch.LongTensor(text_norm)\n",
    "    return text_norm"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## LJ Speech"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hps = utils.get_hparams_from_file(\"./configs/ljs_base.json\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "net_g = SynthesizerTrn(\n",
    "    len(symbols),\n",
    "    hps.data.filter_length // 2 + 1,\n",
    "    hps.train.segment_size // hps.data.hop_length,\n",
    "    **hps.model).cuda()\n",
    "_ = net_g.eval()\n",
    "\n",
    "_ = utils.load_checkpoint(\"/path/to/pretrained_ljs.pth\", net_g, None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "stn_tst = get_text(\"VITS is Awesome!\", hps)\n",
    "with torch.no_grad():\n",
    "    x_tst = stn_tst.cuda().unsqueeze(0)\n",
    "    x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()\n",
    "    audio = net_g.infer(x_tst, x_tst_lengths, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][0,0].data.cpu().float().numpy()\n",
    "ipd.display(ipd.Audio(audio, rate=hps.data.sampling_rate, normalize=False))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## VCTK"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "hps = utils.get_hparams_from_file(\"./configs/vctk_base.json\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "net_g = SynthesizerTrn(\n",
    "    len(symbols),\n",
    "    hps.data.filter_length // 2 + 1,\n",
    "    hps.train.segment_size // hps.data.hop_length,\n",
    "    n_speakers=hps.data.n_speakers,\n",
    "    **hps.model).cuda()\n",
    "_ = net_g.eval()\n",
    "\n",
    "_ = utils.load_checkpoint(\"/path/to/pretrained_vctk.pth\", net_g, None)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "stn_tst = get_text(\"VITS is Awesome!\", hps)\n",
    "with torch.no_grad():\n",
    "    x_tst = stn_tst.cuda().unsqueeze(0)\n",
    "    x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cuda()\n",
    "    sid = torch.LongTensor([4]).cuda()\n",
    "    audio = net_g.infer(x_tst, x_tst_lengths, sid=sid, noise_scale=.667, noise_scale_w=0.8, length_scale=1)[0][0,0].data.cpu().float().numpy()\n",
    "ipd.display(ipd.Audio(audio, rate=hps.data.sampling_rate, normalize=False))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Voice Conversion"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "dataset = TextAudioSpeakerLoader(hps.data.validation_files, hps.data)\n",
    "collate_fn = TextAudioSpeakerCollate()\n",
    "loader = DataLoader(dataset, num_workers=8, shuffle=False,\n",
    "    batch_size=1, pin_memory=True,\n",
    "    drop_last=True, collate_fn=collate_fn)\n",
    "data_list = list(loader)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "with torch.no_grad():\n",
    "    x, x_lengths, spec, spec_lengths, y, y_lengths, sid_src = [x.cuda() for x in data_list[0]]\n",
    "    sid_tgt1 = torch.LongTensor([1]).cuda()\n",
    "    sid_tgt2 = torch.LongTensor([2]).cuda()\n",
    "    sid_tgt3 = torch.LongTensor([4]).cuda()\n",
    "    audio1 = net_g.voice_conversion(spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_tgt1)[0][0,0].data.cpu().float().numpy()\n",
    "    audio2 = net_g.voice_conversion(spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_tgt2)[0][0,0].data.cpu().float().numpy()\n",
    "    audio3 = net_g.voice_conversion(spec, spec_lengths, sid_src=sid_src, sid_tgt=sid_tgt3)[0][0,0].data.cpu().float().numpy()\n",
    "print(\"Original SID: %d\" % sid_src.item())\n",
    "ipd.display(ipd.Audio(y[0].cpu().numpy(), rate=hps.data.sampling_rate, normalize=False))\n",
    "print(\"Converted SID: %d\" % sid_tgt1.item())\n",
    "ipd.display(ipd.Audio(audio1, rate=hps.data.sampling_rate, normalize=False))\n",
    "print(\"Converted SID: %d\" % sid_tgt2.item())\n",
    "ipd.display(ipd.Audio(audio2, rate=hps.data.sampling_rate, normalize=False))\n",
    "print(\"Converted SID: %d\" % sid_tgt3.item())\n",
    "ipd.display(ipd.Audio(audio3, rate=hps.data.sampling_rate, normalize=False))"
   ]
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.7"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}