diff --git a/supervisor/audio.py b/supervisor/audio.py
index d7817662f..0ffd7b9f2 100644
--- a/supervisor/audio.py
+++ b/supervisor/audio.py
@@ -3,12 +3,12 @@ import asyncio
 from contextlib import suppress
 import logging
 from pathlib import Path
-from typing import Awaitable, Optional
 import shutil
+from typing import Awaitable, List, Optional
 
 import jinja2
 
-from .const import ATTR_VERSION, FILE_HASSIO_AUDIO
+from .const import ATTR_VERSION, FILE_HASSIO_AUDIO, STATE_STARTED
 from .coresys import CoreSys, CoreSysAttributes
 from .docker.audio import DockerAudio
 from .docker.stats import DockerStats
@@ -157,6 +157,8 @@ class Audio(JsonConfig, CoreSysAttributes):
             _LOGGER.error("Can't start Audio plugin")
             raise AudioError() from None
 
+        await self._restart_audio_addons()
+
     async def start(self) -> None:
         """Run CoreDNS."""
         # Start Instance
@@ -167,6 +169,8 @@ class Audio(JsonConfig, CoreSysAttributes):
             _LOGGER.error("Can't start Audio plugin")
             raise AudioError() from None
 
+        await self._restart_audio_addons()
+
     def logs(self) -> Awaitable[bytes]:
         """Get CoreDNS docker logs.
 
@@ -217,3 +221,22 @@ class Audio(JsonConfig, CoreSysAttributes):
             default_source=input_profile,
             default_sink=output_profile,
         )
+
+    async def _restart_audio_addons(self):
+        """Restart all Add-ons they can be connect to unix socket."""
+        tasks: List[Awaitable[None]] = []
+
+        # Find all Add-ons running with audio
+        for addon in self.sys_addons.installed:
+            if not addon.with_audio:
+                continue
+            if await addon.state() != STATE_STARTED:
+                continue
+            tasks.append(addon.restart())
+
+        if not tasks:
+            return
+
+        # restart
+        _LOGGER.info("Restart all Add-ons attach to pulse server: %d", len(tasks))
+        await asyncio.wait(tasks)