feat: update emotion handling in audiobook segments and UI for multi-emotion selection

qwen3-tts-backend/core/audiobook_service.py

@@ -16,7 +16,31 @@ from db.models import AudiobookProject, AudiobookCharacter, User
 logger = logging.getLogger(__name__)
 
 _LINE_RE = re.compile(r'^【(.+?)】(.*)$')
-_EMO_RE = re.compile(r'（([^:）]+):([0-9.]+)）\s*$')
+_EMO_RE = re.compile(r'（([^）]+)）\s*$')
+
+
+def _parse_emo(raw: str) -> tuple[Optional[str], Optional[float]]:
+    tokens = [t.strip() for t in raw.split('+') if t.strip()]
+    if not tokens:
+        return None, None
+    weighted = [(':' in t) for t in tokens]
+    if all(weighted) and len(tokens) > 1:
+        return raw, 1.0
+    elif len(tokens) == 1 and weighted[0]:
+        name, _, a = tokens[0].partition(':')
+        try:
+            return name.strip(), float(a)
+        except ValueError:
+            return None, None
+    elif weighted[-1]:
+        last_name, _, a = tokens[-1].rpartition(':')
+        names = tokens[:-1] + [last_name]
+        try:
+            return '+'.join(n.strip() for n in names), float(a)
+        except ValueError:
+            return None, None
+    else:
+        return None, None
 
 # Cancellation events for batch operations, keyed by project_id
 _cancel_events: dict[int, asyncio.Event] = {}
@@ -196,12 +220,10 @@ def parse_ai_script(script_text: str, char_map: dict) -> list[dict]:
         emo_alpha = None
         emo_m = _EMO_RE.search(content)
         if emo_m:
-            emo_text = emo_m.group(1)
-            try:
-                emo_alpha = float(emo_m.group(2))
-            except ValueError:
-                emo_alpha = None
-            content = content[:emo_m.start()].strip()
+            et, ea = _parse_emo(emo_m.group(1))
+            if et is not None:
+                emo_text, emo_alpha = et, ea
+                content = content[:emo_m.start()].strip()
 
         if content.startswith('"') and content.endswith('"'):
             content = content[1:-1].strip()
@@ -211,12 +233,10 @@ def parse_ai_script(script_text: str, char_map: dict) -> list[dict]:
         if emo_text is None:
             emo_m = _EMO_RE.search(content)
             if emo_m:
-                emo_text = emo_m.group(1)
-                try:
-                    emo_alpha = float(emo_m.group(2))
-                except ValueError:
-                    emo_alpha = None
-                content = content[:emo_m.start()].strip()
+                et, ea = _parse_emo(emo_m.group(1))
+                if et is not None:
+                    emo_text, emo_alpha = et, ea
+                    content = content[:emo_m.start()].strip()
 
         character = speaker