fix(inworld): reject malformed base64 TTS audio

ZengWen-DT · ZengWen-DT · commit 56133b13b887 · 2026-07-19T04:15:18.000Z
diff --git a/extensions/inworld/tts.test.ts b/extensions/inworld/tts.test.ts
@@ -237,6 +237,15 @@ describe("inworldTTS", () => {
     );
   });
 
+  it("rejects malformed base64 audio chunks", async () => {
+    const body = JSON.stringify({ result: { audioContent: "not-base64!" } });
+    queueGuardedResponse(new Response(body, { status: 200 }));
+
+    await expect(inworldTTS({ text: "test", apiKey: "fixture-api-key" })).rejects.toThrow(
+      "Inworld TTS returned malformed base64 audio data",
+    );
+  });
+
   it("throws on HTTP errors with response body", async () => {
     queueGuardedResponse(new Response("bad request body", { status: 400 }));
 
diff --git a/extensions/inworld/tts.ts b/extensions/inworld/tts.ts
@@ -1,5 +1,5 @@
 // Inworld plugin module implements tts behavior.
-import { MAX_AUDIO_BYTES } from "openclaw/plugin-sdk/media-runtime";
+import { canonicalizeBase64, MAX_AUDIO_BYTES } from "openclaw/plugin-sdk/media-runtime";
 import { readResponseWithLimit } from "openclaw/plugin-sdk/response-limit-runtime";
 import type { SpeechVoiceOption } from "openclaw/plugin-sdk/speech-core";
 import { fetchWithSsrFGuard, type SsrFPolicy } from "openclaw/plugin-sdk/ssrf-runtime";
@@ -186,7 +186,11 @@ export async function inworldTTS(params: {
       }
 
       if (parsed.result?.audioContent) {
-        const chunk = Buffer.from(parsed.result.audioContent, "base64");
+        const canonicalAudio = canonicalizeBase64(parsed.result.audioContent);
+        if (!canonicalAudio) {
+          throw new Error("Inworld TTS returned malformed base64 audio data");
+        }
+        const chunk = Buffer.from(canonicalAudio, "base64");
         const nextDecodedAudioBytes = decodedAudioBytes + chunk.length;
         if (nextDecodedAudioBytes > MAX_AUDIO_BYTES) {
           throw new Error(
