ipa brace support

styletts2importable.py

@@ -38,6 +38,7 @@ from nltk.tokenize import word_tokenize
 from models import *
 from utils import *
 from text_utils import TextCleaner
+import re
 textclenaer = TextCleaner()
 
 
@@ -136,9 +137,26 @@ sampler = DiffusionSampler(
 
 def inference(text, ref_s, alpha = 0.3, beta = 0.7, diffusion_steps=5, embedding_scale=1, use_gruut=False):
     text = text.strip()
+
+    # search for IPA within []
+    regex = r"\[[^\]]*\]"
+    # remove all non-IPA sections
+    text = text.replace('[]', '')
+    ipa_sections = re.findall(regex, text)
+
+    # replace IPA sections with []
+    if (ipa_sections is not None):
+        text = re.sub(regex, '[]', text, 0, re.MULTILINE)
+        
     ps = global_phonemizer.phonemize([text])
     ps = word_tokenize(ps[0])
     ps = ' '.join(ps)
+
+    # add the IPA back
+    if (ipa_sections is not None):
+        for ipa in ipa_sections:
+            ps = ps.replace('[ ]', ipa, 1)
+            
     tokens = textclenaer(ps)
     tokens.insert(0, 0)
     tokens = torch.LongTensor(tokens).to(device).unsqueeze(0)