[BUG] Inconsistent audio issue

[crashxer]

I know you have your hands full with more important issues right now, but I was trying to find an audio delay setting for my setup today wanted to document my findings somewhere in case you do decide to work on this at some point.

Further down, I attached videos and logs that demonstrate the issue. One of the videos is a screen capture of the catchin' sync app that shows the audio delays precisely.

Describe the bug

I noticed that the audio sync delay doesn't stay consistent and instead can vary from -200ms to +55ms after seeking or pausing/resuming.

A consistent delay isn't a big deal, because it can be compensated easily in an AVR or in the Kodi settings.

But an inconsistent delay that varies in such a large range and can change from negative to positive can't be compensated.

Another user on discord reported that they observed the same issue. So, it's not isolated to my setup.

What do I expect to happen?

Ideally, I would expect the audio delay to remain consistent when the video starts playing, when it is paused+resumed and after seeking.

This way it can be measured and compensated for.

What is actually happening?

The audio sync delay stays consistent as long as you don't seek or pause the video.

But doing one of these actions changes the delay to a very different value, which stays fairly consistent until the next action.

When I initially start playing a video the audio arrives about 100-200ms after the video (in catchin' sync it's a -200ms adjustment).

If I pause and resume, the audio arrives 30-50ms before the video (in catchin' sync it's a +50ms adjustment).

Seeking seems to produce the most inconsistent change in the delay. During most of my test it was -100ms or more. And was consistently negative (audio behind video).

Pausing and resuming was creating a delay in a more narrow range. Frequently in the +40ms to +55ms range (audio ahead of video).

Video of the problem

I used the app catchin' sync, which lets you record a video of an audio sync test file playing on the TV. The app then shows you both the video and an audio waveform. You can align the audio and video to determine the exact audio delay in milliseconds.

In the video I panned up whenever I paused and panned left whenever I skipped backwards to make it easier to scrub through the video and find relevant moments.

• Screen capture of catchin' sync app: https://1drv.ms/v/c/715d84a02a8de423/IQAS--3BhnRMRpI8vxza9DMuATQw-sgN_6piULvFUhrMxYs?e=hPNZRd
• Video of test file playing on the TV: https://1drv.ms/v/c/715d84a02a8de423/IQAgPH5oBPOCRaUYyu84uN7eASupqwJEUnjXs4x_WSmq52c?e=hMXOHe
• Debug Log: https://1drv.ms/u/c/715d84a02a8de423/IQCWFpsZnv7TQYGswJ344kxmAU6_xzx20p276NVYkm7hfwQ?e=5lotEA
  • Debug with Audio, Video and A/V timing components

How to reproduce

You can use the original video of my TV with the Catchin' Sync app and the log to analyze the issue as it happened on my setup.

Or you can reproduce it yourself by following these steps:

1. Download the Catchin' Sync app to precisely measure the audio delay
2. Record a video in Catchin' Sync
   1. Play Spears & Munsil Ulta HD Benchmark (2023) - Disc 1 - 03871 (2160p DV - 23.976fps)
   2. Pause for 10s
   3. Resume and wait another 10-20s
   4. Seek back
   5. Let it play for 10-20s
   6. Pause for a few seconds
   7. Resume for a few seconds
3. Scrub through the video and measure the audio delay after each action

If you don't have the Spears & Munsil disc, you can use one of these free ones, but they don't include DV or TrueHD / Atmos audio:

• PhotoJoseph's free AV sync test files in various FPS
• Catchin' Sync's free AV sync test files
• Salty's Audio Test files

Information

• Core ELEC Version: CoreELEC (p3i): 21.3-Omega_p3i_T3_20260307222442 (Amlogic-ng.arm)
• Kodi Settings:
  • Reset System -> CoreELEC to defaults
    • Audio Fixes mode:: Off
    • Reset PTS offset on seek: On
  • PM4K Alternate Seek: Off
  • Logging: Debug with Audio, Video and A/V timing components
• TV Setup:
  • Streaming Box: Ugoos AM6B+ connected directly to HDMI 1
  • TV: LG C5
  • TV Audio output: Set to TV (no soundbar to isolate the issue)
• Test File:
  • File: Spears & Munsil Ulta HD Benchmark (2023) - Disc 1 - 03871 (2160p DV - 23.976fps)
  • Audio Track: True HD 7.1 with Atmos (First track)

Log file

• Debug Log: https://1drv.ms/u/c/715d84a02a8de423/IQCWFpsZnv7TQYGswJ344kxmAU6_xzx20p276NVYkm7hfwQ?e=5lotEA
  • Debug with Audio, Video and A/V timing components

Context

Workarounds I tried

I tried it with LAV Full and alternate seek enabled, but these had the same inconsistencies.

What kind of solution I am hoping for

I understand that this is a very complex issue and I don't expect the audio delay to be completely gone, but I am hoping it would be possible to make it more consistent.

Perhaps it would be possible to analyze what is happening after pause+resume and try to do the same thing after seeking.

Or perhaps we could try adding an auto-pause after seeking.

Ideally, the delay would stay consistently negative or positive so that at least some compensation could be applied to it.

Thank you so much for considering this issue.

[pannal]

Can you join the Discord thread linked in the README? There is a T3a_dev build that I'd like you to test. I know you :D

[pannal]

Your debug log isn't one. And you've missed component logging.

[crashxer]

I retested on CoreELEC (p3i): 21.3-Omega_p3i_T3a_dev_20260309010039 (Amlogic-ng.arm).

The range has decreased, but the inconsistency is the same:

• At start: -158ms
• After 10s pause: +13ms
• After short delay: +48ms
  • Jumps to -1ms after a few seconds
• After seeking back: -45ms
  • Increased to -85ms just before I paused. Maybe due to OSD?
• After 10s pause: +10ms
• After short pause: +69ms and stayed there until I turned it off

[crashxer]

Another test with debug log:
https://1drv.ms/u/c/715d84a02a8de423/IQAjickkRIRCRJ3WK9LRRfZzAaOuAybbBQQUKkRs51RyohA?e=rBOEfc

• At start: -162
• After 10s pause: +49
• After seeking back: -165
• After 10s pause: +65
• After short pause: +21 and stayed there until I turned it off

[crashxer]

For reference so that it doesn't get lost in discord...

Build: CoreELEC (p3i): 21.3-Omega_p3i_T3a_dev_20260309045814 (Amlogic-ng.arm)

• At start: -139
• After 10s pause: +69
• After short pause: +104, but then quickly recovered to +23
• After seeking back: -69
• After 10s pause: +177, then quickly to +97, then +15ms where it stayed
• After short pause: +XX (didn't write down), then +64 after 2 or 3 seconds, then and stayed there until I turned it off

The multiple measurements after pauses are not changes. Previously I skipped the first few seconds after a change to let it sync up.

LOG:
https://1drv.ms/u/c/715d84a02a8de423/IQBmcUrTaJPQQqOcXAFaCJ1LAY0nFVkSr-FVZPhB4xSDQYw?e=912Q8W

[jakdwaugh]

I know this isn't really part of the bug discussion but can you tell me how you played the Spears & Munsil disc on your ugoos am6b+. Do you have the test pattern as a video file, if you do could you let me know where I could get my hands on such a thing :)
I'm also having a lot of a/v sync issues with the same setup and would like to do some similar testing with catchin sync.

[crashxer]

@jakdwaugh you have to buy the blu rays, rip them and then load them into your media server. The relevant files are on disc 1 and have the filenames 03855.m2ts to 03886.m2ts. I am using the 4k DV 23.976fps test file that is 03871.m2ts.

You can join the discord channel, which is linked to in the readme. Then search for "munsil" there and you will find more information.

[jakdwaugh]

thanks man. I'll look into purchasing the blu ray.

[daffie82]

thanks man. I'll look into purchasing the blu ray.

Hi man, also check here, download links are in the description. These also work great with Catchin' Sync.
https://www.youtube.com/watch?v=c1QgZZ15zDY

[pannal]

You might want to retry with a final build (T3b), or even T4_dev