From 1906dc01bf062195814d346d677c659159f524f3 Mon Sep 17 00:00:00 2001
From: Peter Steinberger <steipete@gmail.com>
Date: Sat, 25 Apr 2026 03:36:54 +0100
Subject: [PATCH] fix(elevenlabs): omit mp3 accept for pcm tts

---
 CHANGELOG.md                      |  1 +
 extensions/elevenlabs/tts.test.ts | 26 ++++++++++++++++++++++++++
 extensions/elevenlabs/tts.ts      | 11 ++++++++++-
 3 files changed, 37 insertions(+), 1 deletion(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 00328d54d22..6f2c09f78f7 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -69,6 +69,7 @@ Docs: https://docs.openclaw.ai
 
 - Dashboard/Windows: open Control UI and OAuth URLs through the system URL handler without `cmd.exe` parsing or PATH-based `rundll32` lookup, and reject non-HTTP browser-open inputs. Fixes #71098. Thanks @Sanjays2402.
 - Providers/OpenAI: separate API-key and Codex sign-in onboarding groups, and avoid replaying stale OpenAI Responses reasoning blocks after a model route switch.
+- Providers/ElevenLabs: omit the MP3-only `Accept` header for PCM telephony synthesis, so Voice Call requests for `pcm_22050` no longer receive MP3 audio. Fixes #67340. Thanks @marcchabot.
 - Skills: honor legacy `metadata.clawdbot` requirements and installer hints when `metadata.openclaw` is absent, so older skills no longer appear ready when required binaries are missing. Fixes #71323. Thanks @chen-zhang-cs-code.
 - Browser/config: expand `~` in `browser.executablePath` before Chromium launch, so home-relative custom browser paths no longer fail with `ENOENT`. Fixes #67264. Thanks @Quratulain-bilal.
 - Telegram/streaming: hide tool-progress status updates by default while keeping explicit `streaming.preview.toolProgress` opt-in support for edited preview messages. Fixes #71320. Thanks @neeravmakwana.
diff --git a/extensions/elevenlabs/tts.test.ts b/extensions/elevenlabs/tts.test.ts
index 65d3accd79c..11042cfd321 100644
--- a/extensions/elevenlabs/tts.test.ts
+++ b/extensions/elevenlabs/tts.test.ts
@@ -24,6 +24,11 @@ describe("elevenlabs tts diagnostics", () => {
     };
   }
 
+  function getHeadersFromFirstFetchCall(fetchMock: ReturnType<typeof vi.fn>): Headers {
+    const init = fetchMock.mock.calls[0]?.[1] as RequestInit | undefined;
+    return new Headers(init?.headers);
+  }
+
   async function expectDefaultTtsRequestToThrow(message: string | RegExp) {
     await expect(elevenLabsTTS(createDefaultTtsRequest())).rejects.toThrow(message);
   }
@@ -80,4 +85,25 @@ describe("elevenlabs tts diagnostics", () => {
 
     expect(streamed.getReadCount()).toBeLessThan(200);
   });
+
+  it("keeps the MPEG Accept header for MP3 output", async () => {
+    const fetchMock = vi.fn(async () => new Response(Buffer.from("mp3")));
+    globalThis.fetch = fetchMock as unknown as typeof fetch;
+
+    await elevenLabsTTS(createDefaultTtsRequest());
+
+    expect(getHeadersFromFirstFetchCall(fetchMock).get("accept")).toBe("audio/mpeg");
+  });
+
+  it("omits the MPEG Accept header for PCM telephony output", async () => {
+    const fetchMock = vi.fn(async () => new Response(Buffer.from("pcm")));
+    globalThis.fetch = fetchMock as unknown as typeof fetch;
+
+    await elevenLabsTTS({
+      ...createDefaultTtsRequest(),
+      outputFormat: "pcm_22050",
+    });
+
+    expect(getHeadersFromFirstFetchCall(fetchMock).has("accept")).toBe(false);
+  });
 });
diff --git a/extensions/elevenlabs/tts.ts b/extensions/elevenlabs/tts.ts
index 42f1d12a3b0..4e81cc1ef93 100644
--- a/extensions/elevenlabs/tts.ts
+++ b/extensions/elevenlabs/tts.ts
@@ -24,6 +24,14 @@ function assertElevenLabsVoiceSettings(settings: {
   requireInRange(settings.speed, 0.5, 2, "speed");
 }
 
+function resolveElevenLabsAcceptHeader(outputFormat: string): string | undefined {
+  const normalized = outputFormat.trim().toLowerCase();
+  if (!normalized || normalized.startsWith("mp3_")) {
+    return "audio/mpeg";
+  }
+  return undefined;
+}
+
 export async function elevenLabsTTS(params: {
   text: string;
   apiKey: string;
@@ -70,6 +78,7 @@ export async function elevenLabsTTS(params: {
   if (outputFormat) {
     url.searchParams.set("output_format", outputFormat);
   }
+  const acceptHeader = resolveElevenLabsAcceptHeader(outputFormat);
 
   const { response, release } = await fetchWithSsrFGuard({
     url: url.toString(),
@@ -78,7 +87,7 @@ export async function elevenLabsTTS(params: {
       headers: {
         "xi-api-key": apiKey,
         "Content-Type": "application/json",
-        Accept: "audio/mpeg",
+        ...(acceptHeader ? { Accept: acceptHeader } : {}),
       },
       body: JSON.stringify({
         text,