English fixed-voice text-to-speech with a Tacotron-lite acoustic model and self-trained HiFi-GAN vocoder.