Tacotron 2 Audio Samples

Audio Samples

Please note that the audio samples are original (without any resampling or other post-processing). So they might not play in Firefox, IE and other browsers that are not supporting WAV format with 16 kHz and 22.05 kHz sampling rates. Please try in Chrome, or download the samples from the GitHub repo located here.

MAILAIBS US was trained using the book “Jane Eyre” read by Elizabeth Klett. MAILAIBS UK was trained using the book “North And South” read by Mary Ann.

Input String LJSpeech MAILAIBS US MAILAIBS UK
I was created by Nvidia’s Deep Learning Software and Research team using the open sequence to sequence framework.
Scientists at the CERN laboratory say they have discovered a new particle.
Generative adversarial network or variational auto-encoder.
Basilar membrane and otolaryngology are not auto-correlations.
He has read the whole thing.
He reads books.
He thought it was time to present the present.
Thisss isrealy awhsome.
He walked all the way home, and he shut the door.
He walked all the way home and he shut the door.
Peter Piper picked a peck of pickled peppers. How many pickled peppers did Peter Piper pick?
She sells sea-shells on the sea-shore. The shells she sells are sea-shells I’m sure.
Tajima Airport serves Toyooka.
This model is a sequence to sequence model consisting of an encoder, attention, decoder network that predicts spectrograms from an input sequence of characters.
quote, that the crowd was about the same as the one which came to see him before but there were one hundred thousand extra people on hand who came to see Misses Kennedy.
and decided that there would be no release of the news of the President’s death until the Vice President had left the hospital.
To meet his liabilities, he raised large sums on forged bills of acceptance drawn upon his mother, a woman of some means,
A court of the collegians was held every Monday to manage its affairs, at which all prisoners were required to attend.
and that, as he was starving, he had resolved on this desperate deed,
Most people would have termed her a splendid woman of her age: and so she was, no doubt, physically speaking;
John Reed was a schoolboy of fourteen years old; four years older than I, for I was but ten: large and stout for his age, with a dingy and unwholesome skin;
Miss Ingram, who had now seated herself with proud grace at the piano, spreading out her snowy robes in queenly amplitude, commenced a brilliant prelude; talking meantime.
triviality, and perhaps imbecility, coarseness, and ill-temper:
The masters-mister Thornton in particular, whose mill had been attacked by Boucher, and who, after the warrant had been issued for his apprehension on the charge of rioting,
Again, stepping nearer, he besought her with another tremulous eager call upon her name.
Candles had been brought, and Fanny had taken up her interminable piece of worsted-work, over which she was yawning;
have been in trade just as much as these Milton-Northern people.
Benefits are in the document.
This is a test.
Its supposed to be nice weather.
It just works.
This is an example of a long snippet of audio that is generated using Taco tron two. Deep learning has advanced multiple fields including but not limited to computer vision, translation, speech recognition, speech synthesis, and more. We hope that it will continue to drive computer science research for the coming years.

* The LJ model tends to scale poorly with long audio sequences. This audio was manually cut.