You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[Bug Report] In-ear detection issue: Video does not pause when removing the left AirPodpro3 #255
Description
Environment:
• App: AirPods Desktop
• Device: AirPodpro3 (Left & Right)
Issue Description:
There is an issue with the in-ear detection for the left AirPodpro3. When I am watching a video and remove the left AirPodpro3, the video fails to pause automatically. However, the in-ear detection on the right AirPodpro3 works completely fine and pauses as expected.
Steps to Reproduce:
Connect AirPodpro3 to the computer and open AirPods Desktop.
Start playing any video.
Remove the left AirPodpro3 from the ear → Video continues playing (Bug).
I looked into this with one pair of AirPods Pro 3 (model 0x2027): connected to Windows 11 with AirPodsDesktop 0.6.1 (--trace), then connected to Linux (BlueZ, btmon + LE scan) for cross-checking. In my case the left bud works and the right one doesn't, the mirror image of this report. What matters seems to be which bud is broadcasting, not left vs. right: in my tests the left bud was the broadcaster almost the whole time.
Short version: AirPodsDesktop's decoding looks correct. My Pro 3 often don't update the advertisement when the non-broadcasting bud is removed, so there is nothing to react to.
Details
Status byte (6th byte of the Apple 07 19 … payload): both in ear = 0x2b. Broadcasting bud out = 0x29; this appeared in every test, and that is the side that works. Other bud out should be 0x23.
On Windows, removing the non-broadcasting (right) bud left the status at 0x2b while advertisements kept arriving about once per second (RSSI −46…−62, rssi_min −100, none dropped). So the RSSI filtering addressed in fix: automatic ear detection on AirPods Pro 2 below the RSSI floor #237 is not the cause here.
On Linux I also opened an AAP session (L2CAP PSM 0x1001, handshake + notification request). AAP reported every removal in every run where it was open (04 00 04 00 06 00 <a> <b>; 0 = in ear, 1 = out, 2 = case). The advertisement mostly did not follow when I repeatedly removed the right bud:
Runs
Playback
Status changes in the advertisement
7
playing continuously
0–2 per run; typically one change, then frozen
2
paused throughout
4 changes over ~6–7 out/in cycles; 2 changes over 5 out/in cycles
2
paused/resumed on every removal (by an AAP-based helper)
every transition (within ~1 s in the run where I timed it)
In one of the "playing" runs the advertisement switched back to 0x2b just as AAP reported the bud coming out, so it can be wrong, not just late.
Whether AAP was open, when it was opened, the buds' state at connection, and whether my other Apple devices had Bluetooth on had no clear effect (one run each, so weak evidence). The only condition where the advertisement tracked every change was when playback was paused and resumed on each removal. My unconfirmed guess is that the non-broadcasting bud's state is refreshed mainly when the audio stream starts or stops; the "paused throughout" runs don't fit that perfectly.
Implication: for my Pro 3 the advertisement is not a reliable source for the non-broadcasting bud. The AAP ear-state packet was reliable, but as far as I know Windows doesn't let user-mode apps open classic-Bluetooth L2CAP channels, so using it would need a kernel driver or a separate Bluetooth stack.
The last 16 bytes of the payload decrypt with the AAP proximity encryption key (AES-128-ECB), but in my captures they only changed when the plaintext part changed, so they don't provide a fresher ear state.
Activity
pibot-42 commentedon Nov 17, 2026
I looked into this with one pair of AirPods Pro 3 (model
0x2027): connected to Windows 11 with AirPodsDesktop 0.6.1 (--trace), then connected to Linux (BlueZ, btmon + LE scan) for cross-checking. In my case the left bud works and the right one doesn't, the mirror image of this report. What matters seems to be which bud is broadcasting, not left vs. right: in my tests the left bud was the broadcaster almost the whole time.Short version: AirPodsDesktop's decoding looks correct. My Pro 3 often don't update the advertisement when the non-broadcasting bud is removed, so there is nothing to react to.
Details
07 19 …payload): both in ear =0x2b. Broadcasting bud out =0x29; this appeared in every test, and that is the side that works. Other bud out should be0x23.0x2bwhile advertisements kept arriving about once per second (RSSI −46…−62,rssi_min−100, none dropped). So the RSSI filtering addressed in fix: automatic ear detection on AirPods Pro 2 below the RSSI floor #237 is not the cause here.0x1001, handshake + notification request). AAP reported every removal in every run where it was open (04 00 04 00 06 00 <a> <b>; 0 = in ear, 1 = out, 2 = case). The advertisement mostly did not follow when I repeatedly removed the right bud:In one of the "playing" runs the advertisement switched back to
0x2bjust as AAP reported the bud coming out, so it can be wrong, not just late.Whether AAP was open, when it was opened, the buds' state at connection, and whether my other Apple devices had Bluetooth on had no clear effect (one run each, so weak evidence). The only condition where the advertisement tracked every change was when playback was paused and resumed on each removal. My unconfirmed guess is that the non-broadcasting bud's state is refreshed mainly when the audio stream starts or stops; the "paused throughout" runs don't fit that perfectly.
Implication: for my Pro 3 the advertisement is not a reliable source for the non-broadcasting bud. The AAP ear-state packet was reliable, but as far as I know Windows doesn't let user-mode apps open classic-Bluetooth L2CAP channels, so using it would need a kernel driver or a separate Bluetooth stack.
Two more notes:
0x23did appear and the broadcasting bud alternated. I can't tell whether that is a model difference or just different conditions.