Avoid implicit TTS clone defaults

Author: skyguan92

internal/openclaw/config.go

@@ -124,7 +124,6 @@ func mergeTTS(cfg map[string]any, managed, next *ManagedState, result *SyncResul
 				"apiKey":  directToolAPIKey(result.APIKey),
 				"baseUrl": result.ProxyAddr,
 				"model":   result.TTSModel.ID,
-				"voice":   "default",
 			},
 		},
 	}

internal/openclaw/plugins/aima-local-tts/index.js

@@ -290,7 +290,6 @@ function resolveConfiguredTTSRoute(cfg) {
     model,
     baseUrl,
     apiKey: asString(provider.apiKey) || defaultLocalAPIKey,
-    voice: asString(provider.voice) || "default",
   };
 }
 
@@ -338,14 +337,19 @@ function buildTTSRequest(route, params, referenceAudioValue, referenceText) {
   if (!allowedResponseFormats.has(responseFormat)) {
     throw new Error(`unsupported response format: ${responseFormat}`);
   }
+  const mode = asString(record.mode);
+  const explicitVoice = readStringParam(record, ["voice"]);
+  const wantsClone = mode === "voice_clone" || asOptionalBoolean(record.xVectorOnlyMode ?? record.x_vector_only_mode, false);
+  if (wantsClone && !referenceAudioValue) {
+    throw new Error("voice cloning requires referenceAudioPath");
+  }
 
   const payload = {
     model: route.model,
     text: readStringParam(record, ["text", "input"]),
-    voice: readStringParam(record, ["voice"]) || route.voice || "default",
     response_format: responseFormat,
-    use_default_reference: asOptionalBoolean(record.useDefaultReference ?? record.use_default_reference, true),
   };
+  if (explicitVoice) payload.voice = explicitVoice;
 
   for (const [source, target] of [
     ["mode", "mode"],
@@ -439,10 +443,6 @@ export default function register(api) {
           type: "boolean",
           description: "Force x-vector-only voice cloning when true.",
         },
-        useDefaultReference: {
-          type: "boolean",
-          description: "Allow the backend to use its configured default reference voice when no reference audio path is supplied.",
-        },
         language: {
           type: "string",
           description: "Optional language hint for the backend.",

internal/openclaw/sync_test.go

@@ -265,6 +265,9 @@ func TestSyncWritesTTSProviderSchema(t *testing.T) {
 	if got := openaiTTS["model"]; got != "qwen3-tts-0.6b" {
 		t.Fatalf("messages.tts.providers.openai.model = %v, want qwen3-tts-0.6b", got)
 	}
+	if _, ok := openaiTTS["voice"]; ok {
+		t.Fatalf("messages.tts.providers.openai.voice should not be written as an implicit default: %v", openaiTTS)
+	}
 	plugins := lookupMap(cfg, "plugins")
 	if plugins == nil {
 		t.Fatal("plugins missing after TTS sync")
@@ -1014,6 +1017,18 @@ func TestDeployPluginsWritesAIMAPlugins(t *testing.T) {
 	if !strings.Contains(string(ttsEntryData), `output path must stay within the OpenClaw workspace`) {
 		t.Fatal("expected TTS plugin to restrict output paths to the OpenClaw workspace")
 	}
+	if strings.Contains(string(ttsEntryData), `use_default_reference`) {
+		t.Fatal("TTS plugin must not ask backends to use a default reference voice")
+	}
+	if strings.Contains(string(ttsEntryData), `useDefaultReference`) {
+		t.Fatal("TTS plugin must not expose a default-reference control")
+	}
+	if strings.Contains(string(ttsEntryData), `route.voice ||`) {
+		t.Fatal("TTS plugin must not treat configured voice as an implicit request default")
+	}
+	if !strings.Contains(string(ttsEntryData), `voice cloning requires referenceAudioPath`) {
+		t.Fatal("expected TTS plugin to require explicit reference audio for voice cloning")
+	}
 
 	if _, err := os.Stat(filepath.Join(targetDir, "aima-local-image", "openclaw.plugin.json")); err != nil {
 		t.Fatalf("expected plugin manifest: %v", err)