fix: Add support for melancholic emotion in text-to-emotion vectors

- The "低落" (melancholic) emotion will always be mapped to "悲伤" (sad) by QwenEmotion's text analysis. It doesn't know the difference between those emotions even if the user writes the exact words.

- Since the words and their meanings are so similar, it might not be possible to train QwenEmotion to learn the difference.

- As a workaround, we perform input text analysis and look for words that mean "melancholic", and swap the "sad" detection result, to make the melancholic/low-energy speech emotion work correctly for users via text-to-emotion.

indextts/infer_v2.py

@@ -614,11 +614,22 @@ class QwenEmotion:
             # TODO: the "低落" (melancholic) emotion will always be mapped to
             # "悲伤" (sad) by QwenEmotion's text analysis. it doesn't know the
             # difference between those emotions even if user writes exact words.
+            # SEE: `self.melancholic_words` for current workaround.
             "低落": "melancholic",
             "惊讶": "surprised",
             "自然": "calm",
         }
         self.desired_vector_order = ["高兴", "愤怒", "悲伤", "恐惧", "反感", "低落", "惊讶", "自然"]
+        self.melancholic_words = {
+            # emotion text phrases that will force QwenEmotion's "悲伤" (sad) detection
+            # to become "低落" (melancholic) instead, to fix limitations mentioned above.
+            "低落",
+            "melancholy",
+            "melancholic",
+            "depression",
+            "depressed",
+            "gloomy",
+        }
         self.max_score = 1.2
         self.min_score = 0.0
 
@@ -686,6 +697,15 @@ class QwenEmotion:
             }
             # print(">> dict result", content)
 
+        # workaround for QwenEmotion's inability to distinguish "悲伤" (sad) vs "低落" (melancholic).
+        # if we detect any of the IndexTTS "melancholic" words, we swap those vectors
+        # to encode the "sad" emotion as "melancholic" (instead of sadness).
+        text_input_lower = text_input.lower()
+        if any(word in text_input_lower for word in self.melancholic_words):
+            # print(">> before vec swap", content)
+            content["悲伤"], content["低落"] = content.get("低落", 0.0), content.get("悲伤", 0.0)
+            # print(">>  after vec swap", content)
+
         return self.convert(content)