support timbre confusion

models/tts/base/tts_inferece.py

@@ -172,9 +172,18 @@ class TTSInference(object):
             out_dir = os.path.join(self.args.output_dir, "single")
             os.makedirs(out_dir, exist_ok=True)
 
-            pred_audio = self.inference_for_single_utterance()
-            save_path = os.path.join(out_dir, "test_pred.wav")
-            save_audio(save_path, pred_audio, self.cfg.preprocess.sample_rate)
+            s1, s2, pred_audio = self.inference_for_single_utterance(length_scale = self.args.duration_control)
+            if s1 is not None and s2 is not None:
+                save_path_s1 = os.path.join(out_dir, "s1.wav")
+                save_path_s2 = os.path.join(out_dir, "s2.wav")
+                save_path_interpolated = os.path.join(out_dir, "test_pred.wav")
+                save_audio(save_path_s1, s1, self.cfg.preprocess.sample_rate)
+                save_audio(save_path_s2, s2, self.cfg.preprocess.sample_rate)
+                save_audio(save_path_interpolated, pred_audio, self.cfg.preprocess.sample_rate)
+            else:
+                save_path_s1 = os.path.join(out_dir, "test_pred.wav")
+                save_audio(save_path_s1, s1, self.cfg.preprocess.sample_rate)
+                
 
         elif self.infer_type == "batch":
             out_dir = os.path.join(self.args.output_dir, "batch")