feat: tts

tts/tts_service.py

@@ -0,0 +1,64 @@
+import io
+import sys
+import time
+
+sys.path.append('tts/vits')
+
+import numpy as np
+import soundfile
+import os
+os.environ["PYTORCH_JIT"] = "0"
+import torch
+
+import tts.vits.commons as commons
+import tts.vits.utils as utils
+
+from tts.vits.models import SynthesizerTrn
+from tts.vits.text.symbols import symbols
+from tts.vits.text import text_to_sequence
+
+import logging
+logging.getLogger().setLevel(logging.INFO)
+logging.basicConfig(level=logging.INFO)
+
+
+from pydub import AudioSegment
+
+
+
+class TTService():
+    
+    def __init__(self, cfg, model, char, speed):
+        logging.info('Initializing TTS Service for %s...' % char)
+        self.hps = utils.get_hparams_from_file(cfg)
+        self.speed = speed
+        self.net_g = SynthesizerTrn(
+            len(symbols),
+            self.hps.data.filter_length // 2 + 1,
+            self.hps.train.segment_size // self.hps.data.hop_length,
+            **self.hps.model).cpu()
+        _ = self.net_g.eval()
+        _ = utils.load_checkpoint(model, self.net_g, None)
+
+    def get_text(self, text, hps):
+        text_norm = text_to_sequence(text, hps.data.text_cleaners)
+        if hps.data.add_blank:
+            text_norm = commons.intersperse(text_norm, 0)
+        text_norm = torch.LongTensor(text_norm)
+        return text_norm
+    
+    def read(self, text, format="wav") -> io.BytesIO:
+        text = text.replace('~', '！')
+        stn_tst = self.get_text(text, self.hps)
+        
+        with torch.no_grad():
+            x_tst = stn_tst.cpu().unsqueeze(0)
+            x_tst_lengths = torch.LongTensor([stn_tst.size(0)]).cpu()
+            # tp = self.net_g.infer(x_tst, x_tst_lengths, noise_scale=.667, noise_scale_w=0.2, length_scale=self.speed)
+            audio = self.net_g.infer(x_tst, x_tst_lengths, noise_scale=.667, noise_scale_w=0.2, length_scale=self.speed)[0][
+                0, 0].data.cpu().float().numpy()
+            f = io.BytesIO()
+            soundfile.write(f, audio, self.hps.data.sampling_rate, format=format)
+            f.seek(0)
+        return f
+