diff --git a/TTS/tts/layers/losses.py b/TTS/tts/layers/losses.py
index 518b7ff3..b506b33f 100644
--- a/TTS/tts/layers/losses.py
+++ b/TTS/tts/layers/losses.py
@@ -512,19 +512,19 @@ class AlignTTSLoss(nn.Module):
         if phase == 0:
             mdn_loss = self.mdn_loss(mu, log_sigma, logp, decoder_target, input_lens, decoder_output_lens)
         elif phase == 1:
-            spec_loss = self.l1(decoder_output, decoder_target, decoder_output_lens)
+            spec_loss = self.spec_loss(decoder_output, decoder_target, decoder_output_lens)
             ssim_loss = self.ssim(decoder_output, decoder_target, decoder_output_lens)
         elif phase == 2:
             mdn_loss = self.mdn_loss(mu, log_sigma, logp, decoder_target, input_lens, decoder_output_lens)
-            spec_loss = self.l1(decoder_output, decoder_target, decoder_output_lens)
+            spec_loss = self.spec_lossX(decoder_output, decoder_target, decoder_output_lens)
             ssim_loss = self.ssim(decoder_output, decoder_target, decoder_output_lens)
         elif phase == 3:
-            dur_loss = self.huber(dur_output, dur_target, input_lens)
+            dur_loss = self.dur_loss(dur_output.unsqueeze(2), dur_target.unsqueeze(2), input_lens)
         else:
             mdn_loss = self.mdn_loss(mu, log_sigma, logp, decoder_target, input_lens, decoder_output_lens)
-            spec_loss = self.l1(decoder_output, decoder_target, decoder_output_lens)
+            spec_loss = self.spec_loss(decoder_output, decoder_target, decoder_output_lens)
             ssim_loss = self.ssim(decoder_output, decoder_target, decoder_output_lens)
-            dur_loss = self.huber(dur_output, dur_target, input_lens)
+            dur_loss = self.dur_loss(dur_output.unsqueeze(2), dur_target.unsqueeze(2), input_lens)
         loss = spec_loss_alpha * spec_loss + ssim_alpha * ssim_loss + dur_loss_alpha * dur_loss + mdn_alpha * mdn_loss
         return {'loss': loss, 'loss_l1': spec_loss, 'loss_ssim': ssim_loss, 'loss_dur': dur_loss, 'mdn_loss': mdn_loss}