diff --git a/apps/desktop/src/app/chat/composer/hooks/use-auto-speak-replies.test.tsx b/apps/desktop/src/app/chat/composer/hooks/use-auto-speak-replies.test.tsx
new file mode 100644
index 0000000000..d8548d1c4c
--- /dev/null
+++ b/apps/desktop/src/app/chat/composer/hooks/use-auto-speak-replies.test.tsx
@@ -0,0 +1,132 @@
+import { act, cleanup, renderHook, waitFor } from '@testing-library/react'
+import { atom } from 'nanostores'
+import { afterEach, describe, expect, it, vi } from 'vitest'
+
+import { assistantTextPart, type ChatMessage, chatMessageText } from '@/lib/chat-messages'
+import { clearSpokenRepliesForTests, markAssistantIdSpoken, resolveSpokenReply } from '@/lib/spoken-reply'
+import { playSpeechText } from '@/lib/voice-playback'
+import { $voicePlayback, setVoicePlaybackState } from '@/store/voice-playback'
+import { $autoSpeakReplies } from '@/store/voice-prefs'
+
+import { ComposerScopeProvider, MAIN_COMPOSER_SCOPE } from '../scope'
+
+import { useAutoSpeakReplies } from './use-auto-speak-replies'
+
+vi.mock('@/lib/voice-playback', () => ({
+  playSpeechText: vi.fn()
+}))
+
+const SESSION_ID = 'session-under-test'
+const IDLE_STATE = { audioElement: null, messageId: null, sequence: 0, source: null, status: 'idle' as const }
+
+function assistantMessage(id: string, text: string): ChatMessage {
+  return { id, parts: [assistantTextPart(text)], role: 'assistant' }
+}
+
+// #93515 — Edge TTS has no chunked-PCM API, so the WS attempt in
+// playSpeechText's fallback ladder settles 'fallback' before any audio plays
+// and the client retries over the POST endpoint. While that POST round-trip
+// is in flight, the backend can rewrite the just-completed reply's renderer
+// id (`assistant-stream-*`) to its durable id. The issue claims
+// `resolveSpokenReply()` fails to follow that rewrite and the reply gets
+// spoken a second time once `$voicePlayback` goes idle.
+describe('useAutoSpeakReplies — Edge TTS fallback chain (#93515)', () => {
+  afterEach(() => {
+    cleanup()
+    clearSpokenRepliesForTests()
+    $autoSpeakReplies.set(false)
+    setVoicePlaybackState({ ...IDLE_STATE })
+    vi.clearAllMocks()
+  })
+
+  it('does not re-speak the reply once playback goes idle after an id rewrite mid-fallback', async () => {
+    $autoSpeakReplies.set(true)
+
+    const $messages = atom<ChatMessage[]>([])
+
+    // The exact pendingReply/markSpoken contract use-composer-voice.ts wires
+    // up for this hook, backed by the real ordinal-anchored dedupe.
+    const pendingReply = () => {
+      const messages = $messages.get()
+      const last = messages.findLast(m => m.role === 'assistant' && !m.hidden)
+      const spoken = resolveSpokenReply(SESSION_ID, messages)
+
+      if (!last || last.id === spoken?.id) {
+        return null
+      }
+
+      return { id: last.id, pending: Boolean(last.pending), text: chatMessageText(last) }
+    }
+
+    const markSpoken = () => {
+      const messages = $messages.get()
+      const last = messages.findLast(m => m.role === 'assistant' && !m.hidden)
+
+      if (last) {
+        markAssistantIdSpoken(SESSION_ID, messages, last.id)
+      }
+    }
+
+    let settleFallback: (() => void) | null = null
+
+    vi.mocked(playSpeechText).mockImplementation(async () => {
+      setVoicePlaybackState({
+        audioElement: null,
+        messageId: 'assistant-stream-1',
+        sequence: 0,
+        source: 'read-aloud',
+        status: 'preparing'
+      })
+
+      // Holds mid-ladder — the WS-fallback-then-POST round trip the issue
+      // describes — until the test rewrites the message id underneath it.
+      await new Promise<void>(resolve => {
+        settleFallback = resolve
+      })
+
+      $messages.set([assistantMessage('durable-42', 'hello there')])
+
+      setVoicePlaybackState({
+        audioElement: null,
+        messageId: 'durable-42',
+        sequence: 0,
+        source: 'read-aloud',
+        status: 'idle'
+      })
+
+      return true
+    })
+
+    renderHook(
+      () =>
+        useAutoSpeakReplies({
+          conversationActive: false,
+          failureLabel: 'read-aloud failed',
+          markSpoken,
+          pendingReply,
+          sessionId: SESSION_ID
+        }),
+      {
+        wrapper: ({ children }) => (
+          <ComposerScopeProvider value={{ ...MAIN_COMPOSER_SCOPE, $messages }}>{children}</ComposerScopeProvider>
+        )
+      }
+    )
+
+    act(() => {
+      $messages.set([assistantMessage('assistant-stream-1', 'hello there')])
+    })
+
+    await waitFor(() => expect(playSpeechText).toHaveBeenCalledTimes(1))
+
+    await act(async () => {
+      settleFallback?.()
+      await Promise.resolve()
+      await Promise.resolve()
+      await Promise.resolve()
+    })
+
+    expect($voicePlayback.get().status).toBe('idle')
+    expect(playSpeechText).toHaveBeenCalledTimes(1)
+  })
+})