From d7c8010c32419f9a3043b07eb0a4ec00a0f4d007 Mon Sep 17 00:00:00 2001
From: allin2 <37659199+allin2@users.noreply.github.com>
Date: Sat, 25 Jul 2026 21:03:34 +0800
Subject: [PATCH 1/2] fix(voice): ignore empty first chunk for speaking event

---
 .../src/voice/__tests__/interrupt-truncation.test.ts  | 11 +++++++++++
 javascript/src/voice/adapter.runtime.ts               |  2 +-
 2 files changed, 12 insertions(+), 1 deletion(-)

diff --git a/javascript/src/voice/__tests__/interrupt-truncation.test.ts b/javascript/src/voice/__tests__/interrupt-truncation.test.ts
index 57c733965..27ce86f31 100644
--- a/javascript/src/voice/__tests__/interrupt-truncation.test.ts
+++ b/javascript/src/voice/__tests__/interrupt-truncation.test.ts
@@ -124,6 +124,17 @@ describe("defaultVoiceCall publishes the agent speaking event", () => {
     await expect(adapter.agentSpeakingEvent!.wait()).resolves.toBeUndefined();
   });
 
+  it("does not set the speaking event when the first chunk is empty", async () => {
+    const adapter = new SpeakingAdapter([
+      new AudioChunk({ data: new Uint8Array(0) }),
+    ]);
+
+    await defaultVoiceCall(adapter, bareInput);
+
+    expect(adapter.agentSpeakingEvent).toBeDefined();
+    expect(adapter.agentSpeakingEvent!.isSet()).toBe(false);
+  });
+
   it("clears the event at the start of the next turn (re-armed)", async () => {
     // Turn-2's first chunk is delayed so the cleared-but-not-yet-set window is
     // observable across a macrotask yield. Without it the immediate chunk would
diff --git a/javascript/src/voice/adapter.runtime.ts b/javascript/src/voice/adapter.runtime.ts
index 09c398fde..03c4e1d07 100644
--- a/javascript/src/voice/adapter.runtime.ts
+++ b/javascript/src/voice/adapter.runtime.ts
@@ -721,8 +721,8 @@ async function drainInner(
   );
   if (first.data.length > 0) {
     onFirstChunk();
+    speakingEvent.set();
   }
-  speakingEvent.set();
 
   const chunks: AudioChunk[] = [first];
   let accumulated = first.durationSeconds;

From 3380f97992847641f75449b012c4eb74f9f1da03 Mon Sep 17 00:00:00 2001
From: allin2 <37659199+allin2@users.noreply.github.com>
Date: Tue, 28 Jul 2026 00:21:41 +0800
Subject: [PATCH 2/2] fix(voice): mark speaking on first playable chunk

---
 .../voice/__tests__/interrupt-truncation.test.ts    | 13 +++++++++++++
 javascript/src/voice/adapter.runtime.ts             | 11 +++++++++--
 2 files changed, 22 insertions(+), 2 deletions(-)

diff --git a/javascript/src/voice/__tests__/interrupt-truncation.test.ts b/javascript/src/voice/__tests__/interrupt-truncation.test.ts
index 27ce86f31..6fa64811f 100644
--- a/javascript/src/voice/__tests__/interrupt-truncation.test.ts
+++ b/javascript/src/voice/__tests__/interrupt-truncation.test.ts
@@ -135,6 +135,19 @@ describe("defaultVoiceCall publishes the agent speaking event", () => {
     expect(adapter.agentSpeakingEvent!.isSet()).toBe(false);
   });
 
+  it("sets the speaking event when audio follows a leading empty chunk", async () => {
+    const adapter = new SpeakingAdapter([
+      new AudioChunk({ data: new Uint8Array(0) }),
+      tone(0.1, "after empty"),
+    ]);
+
+    await defaultVoiceCall(adapter, bareInput);
+
+    expect(adapter.agentSpeakingEvent).toBeDefined();
+    expect(adapter.agentSpeakingEvent!.isSet()).toBe(true);
+    await expect(adapter.agentSpeakingEvent!.wait()).resolves.toBeUndefined();
+  });
+
   it("clears the event at the start of the next turn (re-armed)", async () => {
     // Turn-2's first chunk is delayed so the cleared-but-not-yet-set window is
     // observable across a macrotask yield. Without it the immediate chunk would
diff --git a/javascript/src/voice/adapter.runtime.ts b/javascript/src/voice/adapter.runtime.ts
index 03c4e1d07..a8cf7511b 100644
--- a/javascript/src/voice/adapter.runtime.ts
+++ b/javascript/src/voice/adapter.runtime.ts
@@ -719,10 +719,16 @@ async function drainInner(
     "voice.audio.first_chunk_latency_ms",
     Math.round(performance.now() - receiveStart),
   );
-  if (first.data.length > 0) {
+  // The initial receive can be an empty prefix while the drain still yields
+  // audio. Publish speech on the first playable chunk wherever it appears.
+  let sawPlayableChunk = false;
+  const markFirstPlayableChunk = (chunk: AudioChunk): void => {
+    if (sawPlayableChunk || chunk.data.length === 0) return;
+    sawPlayableChunk = true;
     onFirstChunk();
     speakingEvent.set();
-  }
+  };
+  markFirstPlayableChunk(first);
 
   const chunks: AudioChunk[] = [first];
   let accumulated = first.durationSeconds;
@@ -761,6 +767,7 @@ async function drainInner(
       terminatedReason = "terminal_chunk";
       break;
     }
+    markFirstPlayableChunk(next);
     chunks.push(next);
     accumulated += next.durationSeconds;
     maybeWarnSoftCap();