RO-TransformerTTS+HifiGAN-Vocoder

Added a new blob containing previously used TTS model trained on RSS single
speaker (Adr) + pretrained weights for a HifiGAN vocoder.

TransformerTTS-RO-Adr

TransformerTTS Romanian pretrained model weights.
The model has been trained for 10,000 epochs, using
only one speaker from the RSS dataset (Adr).