diff --git a/backend/cmd/server/main.go b/backend/cmd/server/main.go
index 0b1e8d8..01a9b8e 100644
--- a/backend/cmd/server/main.go
+++ b/backend/cmd/server/main.go
@@ -80,7 +80,7 @@ func main() {
 	}
 
 	// 初始化 Orchestrator
-	orch := orchestrator.New(sttService, llmService, ttsService, sessionMgr, cfg.AI.LLM.Model)
+	orch := orchestrator.New(sttService, llmService, ttsService, sessionMgr, cfg.AI.LLM.Model, cfg.AI.TTS.Voice, cfg.AI.TTS.Speed)
 
 	// Gin 模式
 	if cfg.App.Env == "prod" {
diff --git a/backend/config.yaml b/backend/config.yaml
index d61dc4b..6e427d1 100644
--- a/backend/config.yaml
+++ b/backend/config.yaml
@@ -28,7 +28,7 @@ ai:
   tts:
     provider: mimo
     model: mimo-v2.5-tts
-    voice: 冰糖
+    voice: mimo_default
     speed: 1.0
     endpoint: "https://token-plan-cn.xiaomimimo.com/v1"
     api_key: "tp-c9e7scwfx94qvqyhpnahnw8uaiya01za2qzvg4xe24rp3xiv"
diff --git a/backend/internal/config/config.go b/backend/internal/config/config.go
index 1a5cac8..04f581a 100644
--- a/backend/internal/config/config.go
+++ b/backend/internal/config/config.go
@@ -109,7 +109,7 @@ func Load() (*Config, error) {
 	v.SetDefault("ai.llm.timeout", 10)
 	v.SetDefault("ai.tts.provider", "openai")
 	v.SetDefault("ai.tts.model", "tts-1")
-	v.SetDefault("ai.tts.voice", "alloy")
+	v.SetDefault("ai.tts.voice", "mimo_default")
 	v.SetDefault("ai.tts.speed", 1.0)
 	v.SetDefault("ai.tts.endpoint", "https://api.openai.com/v1")
 	v.SetDefault("ai.tts.timeout", 5)
diff --git a/backend/internal/orchestrator/pipeline.go b/backend/internal/orchestrator/pipeline.go
index d8d2823..fd84c68 100644
--- a/backend/internal/orchestrator/pipeline.go
+++ b/backend/internal/orchestrator/pipeline.go
@@ -23,6 +23,8 @@ type Pipeline struct {
 	ttsService tts.Service
 	sessionMgr session.Manager
 	model      string // LLM 模型名，用于 llm_done 上报
+	ttsVoice   string // TTS 音色
+	ttsSpeed   float64 // TTS 语速
 }
 
 // New 创建 Pipeline 实例。
@@ -32,6 +34,8 @@ func New(
 	ttsService tts.Service,
 	sessionMgr session.Manager,
 	model string,
+	ttsVoice string,
+	ttsSpeed float64,
 ) *Pipeline {
 	return &Pipeline{
 		sttService: sttService,
@@ -39,6 +43,8 @@ func New(
 		ttsService: ttsService,
 		sessionMgr: sessionMgr,
 		model:      model,
+		ttsVoice:   ttsVoice,
+		ttsSpeed:   ttsSpeed,
 	}
 }
 
@@ -306,8 +312,8 @@ func (p *Pipeline) synthesizeTTS(
 	log := logger.Log
 
 	ttsStream, err := p.ttsService.SynthesizeStream(ctx, sentenceCh, tts.Options{
-		Voice:      "alloy",
-		Speed:      1.0,
+		Voice:      p.ttsVoice,
+		Speed:      p.ttsSpeed,
 		OutputFmt:  "mp3",
 		SampleRate: 24000,
 	})