Update app.py

app.py

@@ -48,28 +48,12 @@ def plot_spec_align(mel, align):
     fig_mel = plt.figure()
     ax_mel = fig_mel.add_subplot(111)
     ax_mel.imshow(mel)
-    ax_mel.set_title('Mel-Scale Spectrogram', fontsize=20)
+    ax_mel.set_title('Mel-Scale Spectrogram', fontsize=12)
 
     fig_align = plt.figure()
     ax_align = fig_align.add_subplot(111)
     ax_align.imshow(align)
-    ax_align.set_title('Alignment', fontsize=20)
-
-    '''
-    grid_spec = gridspec.GridSpec(1, 1)
-    ax = plt.subplot(grid_spec[0])
-    plt.imshow(mel)
-    plt.axis('off')
-    ax.set_title('Mel-Scale Spectrogram', fontsize=20)
-
-    ax = plt.subplot(grid_spec[1])
-    plt.imshow(align)
-    plt.axis('off')
-    ax.set_title('Alignment', fontsize=20)
-
-    plt.imshow(legend, interpolation='nearest')
-    plt.grid('off')
-    '''
+    ax_align.set_title('Alignment', fontsize=12)
 
     return fig_mel, fig_align
 
@@ -90,8 +74,9 @@ def synthesize(text, gst_1, gst_2, gst_3):
     audio_numpy = audio.data.cpu().detach().numpy()
 
     # prepare plot for the output:
-    mel_outputs_postnet = mel_outputs_postnet.squeeze().detach().numpy()
-    alignments = alignments.squeeze().detach().numpy()
+    mel_outputs_postnet = torch.flip(mel_outputs_postnet.squeeze(), [0])
+    mel_outputs_postnet = mel_outputs_postnet.detach().numpy()
+    alignments = alignments.squeeze().T.detach().numpy()
     fig_mel, fig_align = plot_spec_align(mel_outputs_postnet, alignments)
 
     return (22050, audio_numpy), fig_mel, fig_align