Merge branch 'dev' of github.com:mozilla/TTS into dev

datasets/preprocess.py

@@ -70,7 +70,7 @@ def mozilla(root_path, meta_file):
     return items
 
 
-def mailabs(root_path, meta_files):
+def mailabs(root_path, meta_files=None):
     """Normalizes M-AI-Labs meta data files to TTS format"""
     speaker_regex = re.compile("by_book/(male|female|mix)/(?P<speaker_name>[^/]+)/")
     if meta_files is None:
@@ -164,4 +164,4 @@ def libri_tts(root_path, meta_files=None):
                 items.append([text, wav_file, speaker_name])
     for item in items:
         assert os.path.exists(item[1]), f" [!] wav file is not exist - {item[1]}"
-    return items
+    return items

utils/synthesis.py

@@ -108,7 +108,7 @@ def synthesis(model,
     # preprocess the given text
     inputs = text_to_seqvec(text, CONFIG, use_cuda)
     speaker_id = id_to_torch(speaker_id)
-    if use_cuda:
+    if speaker_id is not None and use_cuda:
         speaker_id = speaker_id.cuda()
     # synthesize voice
     decoder_output, postnet_output, alignments, stop_tokens = run_model(