diff --git a/system/webrtc/device/video.py b/system/webrtc/device/video.py
index b46d0222030794..8a07c5b9be9c92 100644
--- a/system/webrtc/device/video.py
+++ b/system/webrtc/device/video.py
@@ -1,4 +1,5 @@
 import asyncio
+import struct
 import time
 
 import av
@@ -7,6 +8,13 @@
 from cereal import messaging
 from openpilot.common.realtime import DT_MDL, DT_DMON
 
+# arbitrary 16-byte UUID identifying openpilot frame-timing SEI messages
+TIMING_SEI_UUID = bytes([
+  0xa5, 0xe0, 0xc4, 0xa4, 0x5b, 0x6e, 0x4e, 0x1e,
+  0x9c, 0x7e, 0x12, 0x34, 0x56, 0x78, 0x9a, 0xbc,
+])
+_SEI_PREFIX = b'\x00\x00\x00\x01\x06\x05\x30' + TIMING_SEI_UUID
+
 
 class LiveStreamVideoStreamTrack(TiciVideoStreamTrack):
   camera_to_sock_mapping = {
@@ -22,6 +30,7 @@ def __init__(self, camera_type: str):
     self._sock = self._make_sock(camera_type)
     self._pts = 0
     self._t0_ns = time.monotonic_ns()
+    self.timing_sei_enabled = False
 
   def _make_sock(self, camera_type: str) -> messaging.SubSocket:
     return messaging.sub_sock(self.camera_to_sock_mapping[camera_type], conflate=True)
@@ -29,6 +38,20 @@ def _make_sock(self, camera_type: str) -> messaging.SubSocket:
   def switch_camera(self, camera_type: str) -> None:
     self._sock = self._make_sock(camera_type)
 
+  def _build_frame_data(self, msg) -> bytes:
+    encode_data = getattr(msg, msg.which())
+    if not self.timing_sei_enabled:
+      return encode_data.header + encode_data.data
+
+    idx = encode_data.idx
+    sei_nal = _SEI_PREFIX + struct.pack('>4d',
+      (idx.timestampEof - idx.timestampSof) / 1e6,
+      (msg.logMonoTime - idx.timestampEof) / 1e6,
+      (time.monotonic_ns() - msg.logMonoTime) / 1e6,
+      time.time() * 1000,  # noqa: TID251
+    ) + b'\x80'
+    return encode_data.header + sei_nal + encode_data.data
+
   async def recv(self):
     while True:
       msg = messaging.recv_one_or_none(self._sock)
@@ -36,9 +59,7 @@ async def recv(self):
         break
       await asyncio.sleep(0.005)
 
-    evta = getattr(msg, msg.which())
-
-    packet = av.Packet(evta.header + evta.data)
+    packet = av.Packet(self._build_frame_data(msg))
     packet.time_base = self._time_base
 
     self._pts =  ((time.monotonic_ns() - self._t0_ns) * self._clock_rate) // 1_000_000_000
diff --git a/system/webrtc/webrtcd.py b/system/webrtc/webrtcd.py
index 1239a1759cdea2..502e6716605680 100755
--- a/system/webrtc/webrtcd.py
+++ b/system/webrtc/webrtcd.py
@@ -1,5 +1,6 @@
 #!/usr/bin/env python3
 
+import time
 import argparse
 import asyncio
 import contextlib
@@ -173,12 +174,30 @@ async def get_answer(self):
   def message_handler(self, message: bytes):
     assert self.incoming_bridge is not None
     try:
-      msg_json = json.loads(message)
-      if msg_json.get("type") == "livestreamCameraSwitch" and hasattr(self.video_track, "switch_camera"):
-        self.video_track.switch_camera(msg_json["data"]["camera"])
-        return
-
-      if msg_json.get("type") not in self.incoming_bridge_services:
+      payload = json.loads(message) if isinstance(message, (bytes, str)) else None
+      if isinstance(payload, dict):
+        msg_type = payload.get("type")
+
+        if msg_type == "livestreamCameraSwitch":
+          self.video_track.switch_camera(payload["data"]["camera"])
+          return
+
+        if msg_type == "clockSync":
+          data = payload.get("data", {})
+          pong = json.dumps({"type": "clockSync", "data": {
+            "action": "pong", "browserSendTime": data.get("browserSendTime"), "deviceTime": time.time() * 1000, # noqa: TID251
+          }})
+          self.stream.get_messaging_channel().send(pong)
+          return
+
+        if msg_type == "enableTimingSei":
+          enabled = bool(payload.get("data", {}).get("enabled"))
+          for track in self.video_tracks:
+            if hasattr(track, 'timing_sei_enabled'):
+              track.timing_sei_enabled = enabled
+          return
+
+      if payload.get("type") not in self.incoming_bridge_services:
         return
       self.incoming_bridge.send(message)
     except Exception: