Release 0.4.1.2

Fixed

• Text-to-speech requests run asynchronously, closing a race that dropped or doubled speech
• Base64 decoding of text-to-speech audio is fixed