Add arxiv link for Wave-Tacotron paper.

index.html

@@ -223,5 +223,16 @@ <h2>Publications</h2>
         </ul>
       </header>
     </article>
+
+    <article>
+      <header>
+        <span class="paper_date">(November 2020)</span>
+        <span class="paper_title">Wave-Tacotron: Spectrogram-free end-to-end text-to-speech synthesis</span>
+        <ul>
+          <li><a href="https://arxiv.org/abs/2011.03568">paper</a></li>
+          <li><a href="publications/wave-tacotron/index.html">audio samples</a></li>
+        </ul>
+      </header>
+    </article>
   </body>
 </html>

publications/wave-tacotron/index.html

@@ -88,12 +88,7 @@
         <h1>Audio samples from "Wave-Tacotron: Spectrogram-free end-to-end text-to-speech synthesis"</h1>
       </header>
     </article>
-    <!--
-      <div><p><b>Paper:</b> <a href="">arXiv</a></p></div>
-    -->
-    <!--
-    <div><p><b>Paper:</b> <a href="">arXiv</a></p></div>
-    -->
+    <div><p><b>Paper:</b> <a href="https://arxiv.org/abs/2011.03568">arXiv</a></p></div>
     <div><p><b>Authors:</b> Ron J. Weiss, RJ Skerry-Ryan, Eric Battenberg, Soroosh Mariooryad, Diederik P. Kingma</p></div>
     <div><p><b>Abstract:</b> We describe a sequence-to-sequence neural network which can directly generate speech waveforms from text inputs. The architecture extends the Tacotron model by incorporating a normalizing flow into the autoregressive decoder loop. Output waveforms are modeled as a sequence of non-overlapping fixed-length frames, each one containing hundreds of samples. The interdependencies of waveform samples within each frame are modeled using the normalizing flow, enabling parallel training and synthesis. Longer-term dependencies are handled autoregressively by conditioning each flow on preceding frames. This model can be optimized directly with maximum likelihood, without using intermediate, hand-designed features nor additional loss terms. Contemporary state-of-the-art text-to-speech (TTS) systems use a cascade of separately learned models: one (such as Tacotron) which generates intermediate features (such as spectrograms) from text, followed by a vocoder (such as WaveRNN) which generates waveform samples from the intermediate features.The proposed system, in contrast, does not use a fixed intermediate representation, and learns all parameters end-to-end. Experiments show that the proposed model generates speech with quality approaching a state-of-the-art neural TTS system, with significantly improved generation speed.</p></div>