feat(tts): parse MiniMax voice gender + language from naming convention

MiniMax /v1/get_voice does not return explicit gender/language fields,
but the system-voice naming convention exposes them:

- voice_id prefix `male-*` / `female-*` (legacy Chinese voices)
- voice_name suffix `*_Man` / `*_Boy` / `*_Lord` / `*_Speaker` → male
- voice_name suffix `*_Lady` / `*_Girl` / `*_Belle` / `*_Lass` → female
- voice_name prefix `English_*` / `Chinese_*` / `Japanese_*` /
  `Korean_*` / `French_*` / etc. → language

`parseMinimaxLabels` is best-effort — returns nil for unparseable IDs
(e.g. `moss_audio_<uuid>` cloned voices), so the picker shows no badge
rather than a wrong one. The picker already renders `voice.labels` as
badges; nothing to wire on the FE side.

8-case unit test covers the matrix.

internal/audio/minimax/voices.go

@@ -7,6 +7,7 @@ import (
 	"fmt"
 	"io"
 	"net/http"
+	"strings"
 	"sync"
 	"time"
 
@@ -136,6 +137,7 @@ func (vl *VoiceLister) fetchVoices(ctx context.Context) ([]audio.Voice, error) {
 			ID:       v.VoiceID,
 			Name:     v.VoiceName,
 			Category: "system",
+			Labels:   parseMinimaxLabels(v.VoiceID, v.VoiceName),
 		})
 	}
 	for _, v := range result.VoiceCloning {
@@ -154,3 +156,49 @@ func (vl *VoiceLister) fetchVoices(ctx context.Context) ([]audio.Voice, error) {
 	}
 	return voices, nil
 }
+
+// parseMinimaxLabels extracts gender + language hints from MiniMax system voice
+// IDs and names. MiniMax API does not return these fields explicitly, but the
+// naming convention exposes them:
+//   - voice_id prefix `male-*` / `female-*` (legacy Chinese voices)
+//   - voice_name suffix `*_Man` / `*_Boy` / `*_Lord` → male
+//   - voice_name suffix `*_Lady` / `*_Girl` / `*_Belle` / `*_Lass` → female
+//   - voice_name prefix `English_*` / `Chinese_*` / `Japanese_*` / `Korean_*` →
+//     language. Returns nil when nothing parseable, so the picker shows no
+//     badge instead of an empty one.
+func parseMinimaxLabels(voiceID, voiceName string) map[string]string {
+	out := map[string]string{}
+	idLower := strings.ToLower(voiceID)
+	nameLower := strings.ToLower(voiceName)
+
+	switch {
+	case strings.HasPrefix(idLower, "male-") ||
+		hasAnySuffix(nameLower, "_man", "_boy", "_lord", "_speaker", "_uncle"):
+		out["gender"] = "male"
+	case strings.HasPrefix(idLower, "female-") ||
+		hasAnySuffix(nameLower, "_lady", "_girl", "_belle", "_lass", "_woman", "_aunt"):
+		out["gender"] = "female"
+	}
+
+	for _, lang := range []string{"english", "chinese", "japanese", "korean", "french", "spanish", "german", "italian", "portuguese"} {
+		if strings.HasPrefix(nameLower, lang+"_") {
+			// Capitalize first rune; lang labels are ASCII so no need for x/text/cases.
+			out["language"] = strings.ToUpper(lang[:1]) + lang[1:]
+			break
+		}
+	}
+
+	if len(out) == 0 {
+		return nil
+	}
+	return out
+}
+
+func hasAnySuffix(s string, suffixes ...string) bool {
+	for _, suf := range suffixes {
+		if strings.HasSuffix(s, suf) {
+			return true
+		}
+	}
+	return false
+}

internal/audio/minimax/voices_internal_test.go

@@ -0,0 +1,37 @@
+package minimax
+
+import "testing"
+
+// TestParseMinimaxLabels covers the heuristic that derives gender + language
+// hints from MiniMax voice IDs/names. MiniMax API does not return these fields
+// explicitly, so we depend on the documented naming convention.
+func TestParseMinimaxLabels(t *testing.T) {
+	tests := []struct {
+		name      string
+		voiceID   string
+		voiceName string
+		want      map[string]string
+	}{
+		{"male prefix", "male-qn-qingse", "male-qn-qingse", map[string]string{"gender": "male"}},
+		{"female prefix", "female-tianmei", "female-tianmei", map[string]string{"gender": "female"}},
+		{"english man suffix", "abc", "English_Persuasive_Man", map[string]string{"gender": "male", "language": "English"}},
+		{"english lady suffix", "xyz", "English_Graceful_Lady", map[string]string{"gender": "female", "language": "English"}},
+		{"english girl suffix", "xyz", "English_radiant_girl", map[string]string{"gender": "female", "language": "English"}},
+		{"japanese belle suffix", "abc", "Japanese_Whisper_Belle", map[string]string{"gender": "female", "language": "Japanese"}},
+		{"unknown name no prefix", "moss_audio_xyz", "moss_audio_xyz", nil},
+		{"empty inputs", "", "", nil},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			got := parseMinimaxLabels(tc.voiceID, tc.voiceName)
+			if len(got) != len(tc.want) {
+				t.Fatalf("len mismatch: got %v, want %v", got, tc.want)
+			}
+			for k, v := range tc.want {
+				if got[k] != v {
+					t.Errorf("key %q: got %q, want %q", k, got[k], v)
+				}
+			}
+		})
+	}
+}

internal/audio/minimax/voices_test.go

@@ -178,3 +178,4 @@ func TestListVoices_5xxFallback_ReturnsStaleCacheWhenPresent(t *testing.T) {
 		t.Errorf("second call (cached): expected 2 voices, got %d", len(voices2))
 	}
 }
+