I have trained Tacotron-1 a model via LJ,Speech database, the synthesis quality is very good but I rarely get synthesis problems for some test inputs. For example:
for the text:
“the company behind multiplayer games including World of Warcraft and Overwatch, reported its service had suffered as a result of a cyber attack”
TTS did not synthesize after the word"Warcraft".
If there is anyone having trained a Tacotron-1 via LJSpeech recently, can you try the same text if it synthesize the whole text correctly?