fix(review): preserve talk directive overrides

obviyus · obviyus · commit 47e412bd0b2b · 2026-03-20T11:01:24.000+05:30
diff --git a/apps/android/app/src/main/java/ai/openclaw/app/voice/TalkModeManager.kt b/apps/android/app/src/main/java/ai/openclaw/app/voice/TalkModeManager.kt
@@ -723,6 +723,9 @@ class TalkModeManager(
         TalkModeRuntime.validatedLanguage(directive?.language)?.let {
           put("language", JsonPrimitive(it))
         }
+        directive?.outputFormat?.trim()?.takeIf { it.isNotEmpty() }?.let {
+          put("outputFormat", JsonPrimitive(it))
+        }
       }
     val res = session.request("talk.speak", params.toString())
     val root = json.parseToJsonElement(res).asObjectOrNull() ?: error("talk.speak returned invalid JSON")
diff --git a/src/gateway/protocol/schema/channels.ts b/src/gateway/protocol/schema/channels.ts
@@ -21,6 +21,7 @@ export const TalkSpeakParamsSchema = Type.Object(
     text: NonEmptyString,
     voiceId: Type.Optional(Type.String()),
     modelId: Type.Optional(Type.String()),
+    outputFormat: Type.Optional(Type.String()),
     speed: Type.Optional(Type.Number()),
     stability: Type.Optional(Type.Number()),
     similarity: Type.Optional(Type.Number()),
diff --git a/src/gateway/server-methods/talk.ts b/src/gateway/server-methods/talk.ts
@@ -69,7 +69,13 @@ function resolveTalkVoiceId(
   if (!aliases) {
     return requested;
   }
-  return aliases[normalizeAliasKey(requested)] ?? requested;
+  const normalizedRequested = normalizeAliasKey(requested);
+  for (const [alias, voiceId] of Object.entries(aliases)) {
+    if (normalizeAliasKey(alias) === normalizedRequested) {
+      return voiceId;
+    }
+  }
+  return requested;
 }
 
 function readTalkVoiceSettings(
@@ -189,6 +195,7 @@ function buildTalkSpeakOverrides(
 ): TtsDirectiveOverrides {
   const voiceId = resolveTalkVoiceId(providerConfig, trimString(params.voiceId));
   const modelId = trimString(params.modelId);
+  const outputFormat = trimString(params.outputFormat);
   const speed = finiteNumber(params.speed);
   const seed = finiteNumber(params.seed);
   const normalize = normalizeTextNormalization(params.normalize);
@@ -212,6 +219,7 @@ function buildTalkSpeakOverrides(
     overrides.elevenlabs = {
       ...(voiceId == null ? {} : { voiceId }),
       ...(modelId == null ? {} : { modelId }),
+      ...(outputFormat == null ? {} : { outputFormat }),
       ...(seed == null ? {} : { seed }),
       ...(normalize == null ? {} : { applyTextNormalization: normalize }),
       ...(language == null ? {} : { languageCode: language }),
@@ -230,7 +238,10 @@ function buildTalkSpeakOverrides(
   }
 
   if (provider === "microsoft") {
-    overrides.microsoft = voiceId == null ? undefined : { voice: voiceId };
+    overrides.microsoft = {
+      ...(voiceId == null ? {} : { voice: voiceId }),
+      ...(outputFormat == null ? {} : { outputFormat }),
+    };
   }
 
   return overrides;
diff --git a/src/gateway/server.talk-config.test.ts b/src/gateway/server.talk-config.test.ts
@@ -301,4 +301,51 @@ describe("gateway talk.config", () => {
       globalThis.fetch = originalFetch;
     }
   });
+
+  it("resolves talk voice aliases case-insensitively and forwards output format", async () => {
+    const { writeConfigFile } = await import("../config/config.js");
+    await writeConfigFile({
+      talk: {
+        provider: "elevenlabs",
+        providers: {
+          elevenlabs: {
+            apiKey: "elevenlabs-talk-key", // pragma: allowlist secret
+            voiceId: "voice-default",
+            voiceAliases: {
+              Clawd: "EXAVITQu4vr4xnSDxMaL",
+            },
+          },
+        },
+      },
+    });
+
+    const originalFetch = globalThis.fetch;
+    let fetchUrl: string | undefined;
+    const fetchMock = vi.fn(async (input: RequestInfo | URL) => {
+      fetchUrl = typeof input === "string" ? input : input instanceof URL ? input.href : input.url;
+      return new Response(new Uint8Array([4, 5, 6]), { status: 200 });
+    });
+    globalThis.fetch = fetchMock as typeof fetch;
+
+    try {
+      await withServer(async (ws) => {
+        await connectOperator(ws, ["operator.read", "operator.write"]);
+        const res = await fetchTalkSpeak(ws, {
+          text: "Hello from talk mode.",
+          voiceId: "clawd",
+          outputFormat: "pcm_44100",
+        });
+        expect(res.ok).toBe(true);
+        expect(res.payload?.provider).toBe("elevenlabs");
+        expect(res.payload?.outputFormat).toBe("pcm_44100");
+        expect(res.payload?.audioBase64).toBe(Buffer.from([4, 5, 6]).toString("base64"));
+      });
+
+      expect(fetchMock).toHaveBeenCalled();
+      expect(fetchUrl).toContain("/v1/text-to-speech/EXAVITQu4vr4xnSDxMaL");
+      expect(fetchUrl).toContain("output_format=pcm_44100");
+    } finally {
+      globalThis.fetch = originalFetch;
+    }
+  });
 });
diff --git a/src/tts/providers/elevenlabs.ts b/src/tts/providers/elevenlabs.ts
@@ -72,7 +72,9 @@ export function buildElevenLabsSpeechProvider(): SpeechProviderPlugin {
       if (!apiKey) {
         throw new Error("ElevenLabs API key missing");
       }
-      const outputFormat = req.target === "voice-note" ? "opus_48000_64" : "mp3_44100_128";
+      const outputFormat =
+        req.overrides?.elevenlabs?.outputFormat ??
+        (req.target === "voice-note" ? "opus_48000_64" : "mp3_44100_128");
       const audioBuffer = await elevenLabsTTS({
         text: req.text,
         apiKey,
diff --git a/src/tts/providers/microsoft.ts b/src/tts/providers/microsoft.ts
@@ -83,7 +83,7 @@ export function buildMicrosoftSpeechProvider(): SpeechProviderPlugin {
       const tempRoot = resolvePreferredOpenClawTmpDir();
       mkdirSync(tempRoot, { recursive: true, mode: 0o700 });
       const tempDir = mkdtempSync(path.join(tempRoot, "tts-microsoft-"));
-      let outputFormat = req.config.edge.outputFormat;
+      let outputFormat = req.overrides?.microsoft?.outputFormat ?? req.config.edge.outputFormat;
       const fallbackOutputFormat =
         outputFormat !== DEFAULT_EDGE_OUTPUT_FORMAT ? DEFAULT_EDGE_OUTPUT_FORMAT : undefined;
 
diff --git a/src/tts/tts.ts b/src/tts/tts.ts
@@ -167,13 +167,15 @@ export type TtsDirectiveOverrides = {
   elevenlabs?: {
     voiceId?: string;
     modelId?: string;
+    outputFormat?: string;
     seed?: number;
     applyTextNormalization?: "auto" | "on" | "off";
     languageCode?: string;
     voiceSettings?: Partial<ResolvedTtsConfig["elevenlabs"]["voiceSettings"]>;
   };
   microsoft?: {
     voice?: string;
+    outputFormat?: string;
   };
 };
 

Original file line number	Diff line number	Diff line change
`@@ -723,6 +723,9 @@ class TalkModeManager(`
`723`	`723`	`TalkModeRuntime.validatedLanguage(directive?.language)?.let {`
`724`	`724`	`put("language", JsonPrimitive(it))`
`725`	`725`	`}`
	`726`	`+ directive?.outputFormat?.trim()?.takeIf { it.isNotEmpty() }?.let {`
	`727`	`+ put("outputFormat", JsonPrimitive(it))`
	`728`	`+ }`
`726`	`729`	`}`
`727`	`730`	`val res = session.request("talk.speak", params.toString())`
`728`	`731`	`val root = json.parseToJsonElement(res).asObjectOrNull() ?: error("talk.speak returned invalid JSON")`