| import numpy as np |
| import torch |
|
|
| from bert_vits2 import commons |
| from bert_vits2 import utils as bert_vits2_utils |
| from bert_vits2.models import SynthesizerTrn |
| from bert_vits2.text import * |
| from bert_vits2.text.cleaner import clean_text |
| from bert_vits2.utils import process_legacy_versions |
| from contants import ModelType |
| from utils import classify_language, get_hparams_from_file, lang_dict |
| from utils.sentence import sentence_split_and_markup, cut |
|
|
|
|
| class Bert_VITS2: |
| def __init__(self, model, config, device=torch.device("cpu"), **kwargs): |
| self.hps_ms = get_hparams_from_file(config) if isinstance(config, str) else config |
| self.n_speakers = getattr(self.hps_ms.data, 'n_speakers', 0) |
| self.speakers = [item[0] for item in |
| sorted(list(getattr(self.hps_ms.data, 'spk2id', {'0': 0}).items()), key=lambda x: x[1])] |
| self.symbols = symbols |
| |
|
|
| |
| self.version = process_legacy_versions(self.hps_ms) |
|
|
| if self.version in ["1.0", "1.0.0", "1.0.1"]: |
| self.symbols = symbols_legacy |
| self.hps_ms.model.n_layers_trans_flow = 3 |
| |
|
|
| elif self.version in ["1.1.0-transition"]: |
| self.hps_ms.model.n_layers_trans_flow = 3 |
|
|
| elif self.version in ["1.1", "1.1.0", "1.1.1"]: |
| self.hps_ms.model.n_layers_trans_flow = 6 |
|
|
| key = f"{ModelType.BERT_VITS2.value}_v{self.version}" if self.version else ModelType.BERT_VITS2.value |
| self.lang = lang_dict.get(key, ["unknown"]) |
| self.bert_handler = BertHandler(self.lang) |
|
|
| self._symbol_to_id = {s: i for i, s in enumerate(self.symbols)} |
|
|
| self.net_g = SynthesizerTrn( |
| len(self.symbols), |
| self.hps_ms.data.filter_length // 2 + 1, |
| self.hps_ms.train.segment_size // self.hps_ms.data.hop_length, |
| n_speakers=self.hps_ms.data.n_speakers, |
| symbols=self.symbols, |
| **self.hps_ms.model).to(device) |
| _ = self.net_g.eval() |
| self.device = device |
| self.load_model(model) |
|
|
| def load_model(self, model): |
| bert_vits2_utils.load_checkpoint(model, self.net_g, None, skip_optimizer=True, version=self.version) |
|
|
| def get_speakers(self): |
| return self.speakers |
|
|
| @property |
| def sampling_rate(self): |
| return self.hps_ms.data.sampling_rate |
|
|
| def get_text(self, text, language_str, hps): |
| norm_text, phone, tone, word2ph = clean_text(text, language_str) |
| phone, tone, language = cleaned_text_to_sequence(phone, tone, language_str, self._symbol_to_id) |
|
|
| if hps.data.add_blank: |
| phone = commons.intersperse(phone, 0) |
| tone = commons.intersperse(tone, 0) |
| language = commons.intersperse(language, 0) |
| for i in range(len(word2ph)): |
| word2ph[i] = word2ph[i] * 2 |
| word2ph[0] += 1 |
| bert = self.bert_handler.get_bert(norm_text, word2ph, language_str) |
| del word2ph |
| assert bert.shape[-1] == len(phone), phone |
|
|
| if language_str == "zh": |
| bert = bert |
| ja_bert = torch.zeros(768, len(phone)) |
| elif language_str == "ja": |
| ja_bert = bert |
| bert = torch.zeros(1024, len(phone)) |
| else: |
| bert = torch.zeros(1024, len(phone)) |
| ja_bert = torch.zeros(768, len(phone)) |
| assert bert.shape[-1] == len( |
| phone |
| ), f"Bert seq len {bert.shape[-1]} != {len(phone)}" |
| phone = torch.LongTensor(phone) |
| tone = torch.LongTensor(tone) |
| language = torch.LongTensor(language) |
| return bert, ja_bert, phone, tone, language |
|
|
| def infer(self, text, lang, sdp_ratio, noise_scale, noise_scale_w, length_scale, sid): |
| bert, ja_bert, phones, tones, lang_ids = self.get_text(text, lang, self.hps_ms) |
| with torch.no_grad(): |
| x_tst = phones.to(self.device).unsqueeze(0) |
| tones = tones.to(self.device).unsqueeze(0) |
| lang_ids = lang_ids.to(self.device).unsqueeze(0) |
| bert = bert.to(self.device).unsqueeze(0) |
| ja_bert = ja_bert.to(self.device).unsqueeze(0) |
| x_tst_lengths = torch.LongTensor([phones.size(0)]).to(self.device) |
| speakers = torch.LongTensor([int(sid)]).to(self.device) |
| audio = self.net_g.infer(x_tst, x_tst_lengths, speakers, tones, lang_ids, bert, ja_bert, sdp_ratio=sdp_ratio |
| , noise_scale=noise_scale, noise_scale_w=noise_scale_w, length_scale=length_scale)[ |
| 0][0, 0].data.cpu().float().numpy() |
|
|
| torch.cuda.empty_cache() |
| return audio |
|
|
| def get_audio(self, voice, auto_break=False): |
| text = voice.get("text", None) |
| lang = voice.get("lang", "auto") |
| sdp_ratio = voice.get("sdp_ratio", 0.2) |
| noise_scale = voice.get("noise", 0.5) |
| noise_scale_w = voice.get("noisew", 0.6) |
| length_scale = voice.get("length", 1) |
| sid = voice.get("id", 0) |
| max = voice.get("max", 50) |
| |
| if lang == "auto": |
| lang = classify_language(text, target_languages=self.lang) |
|
|
| sentence_list = cut(text, max) |
| audios = [] |
| for sentence in sentence_list: |
| audio = self.infer(sentence, lang, sdp_ratio, noise_scale, noise_scale_w, length_scale, sid) |
| audios.append(audio) |
| audio = np.concatenate(audios) |
| return audio |
|
|
