fix(auto-reply): preserve silent voice payloads

CHANGELOG.md

@@ -14,6 +14,7 @@ Docs: https://docs.openclaw.ai
 ### Fixes
 
 - Auto-reply/commands: stop bare `/reset` and `/new` after reset hooks acknowledge the command, so non-ACP channels no longer fall through into empty provider calls while `/reset <message>` and `/new <message>` still seed the next model turn. Fixes #73367. Thanks @hoyanhan and @wenxu007.
+- Auto-reply: preserve voice-note media from silent turns while continuing to suppress text and non-voice media, so `NO_REPLY` TTS replies still deliver the requested audio bubble. (#73406) Thanks @zqchris.
 - Agents/Anthropic: send implicit Anthropic beta headers only to direct public Anthropic endpoints, including OAuth, so custom Anthropic-compatible providers no longer mis-handle unsupported beta flags unless explicitly configured. Refs #73346. Thanks @byBrodowski.
 - Skills: require explicit `skills.entries.coding-agent.enabled` before exposing the bundled coding-agent skill, so installs with Codex on PATH but no OpenAI auth do not silently offer Codex delegation. Fixes #73358. Thanks @LaFleurAdvertising and @Sanjays2402.
 - Plugins/startup: precompute bundled runtime mirror fingerprints before taking the mirror lock and keep Docker bundled plugin runtime deps/mirrors in a Docker-managed volume instead of the Windows/WSL config bind mount, so cold starts avoid slow host-volume mirror writes. Fixes #73339. Thanks @1yihui.

src/auto-reply/reply/agent-runner-payloads.test.ts

@@ -371,7 +371,7 @@ describe("buildReplyPayloads media filter integration", () => {
     });
   });
 
-  it("drops all final payloads during silent turns, including media-only payloads", async () => {
+  it("drops non-voice final payloads during silent turns, including media-only payloads", async () => {
     const { replyPayloads } = await buildReplyPayloads({
       ...baseParams,
       silentExpected: true,
@@ -381,6 +381,31 @@ describe("buildReplyPayloads media filter integration", () => {
     expect(replyPayloads).toHaveLength(0);
   });
 
+  it("keeps voice media payloads during silent turns", async () => {
+    const { replyPayloads } = await buildReplyPayloads({
+      ...baseParams,
+      silentExpected: true,
+      payloads: [{ text: "NO_REPLY", mediaUrl: "file:///tmp/voice.opus", audioAsVoice: true }],
+    });
+
+    expect(replyPayloads).toHaveLength(1);
+    expect(replyPayloads[0]).toMatchObject({
+      text: undefined,
+      mediaUrl: "file:///tmp/voice.opus",
+      audioAsVoice: true,
+    });
+  });
+
+  it("drops empty voice markers during silent turns", async () => {
+    const { replyPayloads } = await buildReplyPayloads({
+      ...baseParams,
+      silentExpected: true,
+      payloads: [{ audioAsVoice: true }],
+    });
+
+    expect(replyPayloads).toHaveLength(0);
+  });
+
   it("suppresses warning text when silent media payloads fail normalization", async () => {
     const normalizeMediaPaths = async () => {
       throw new Error("file not found");

src/auto-reply/reply/agent-runner-payloads.ts

@@ -87,6 +87,10 @@ async function normalizeSentMediaUrlsForDedupe(params: {
   return normalizedUrls;
 }
 
+function shouldKeepPayloadDuringSilentTurn(payload: ReplyPayload): boolean {
+  return payload.audioAsVoice === true && resolveSendableOutboundReplyParts(payload).hasMedia;
+}
+
 export async function buildReplyPayloads(params: {
   payloads: ReplyPayload[];
   isHeartbeat: boolean;
@@ -165,7 +169,9 @@ export async function buildReplyPayloads(params: {
       }),
     )
   ).filter(isRenderablePayload);
-  const silentFilteredPayloads = params.silentExpected ? [] : replyTaggedPayloads;
+  const silentFilteredPayloads = params.silentExpected
+    ? replyTaggedPayloads.filter(shouldKeepPayloadDuringSilentTurn)
+    : replyTaggedPayloads;
 
   // Drop final payloads only when block streaming succeeded end-to-end.
   // If streaming aborted (e.g., timeout), fall back to final payloads.