[wizziwig]

Seriously? 1/4 of a second would be unwatchable for me. That's around 6 frames at 24p. Once you get to that level it starts to look like you are watching a dubbed movie. Most people speak at around 3 words per second, so you are almost a word out of sync.

Agreed. No idea how anyone could tolerate that if those measured a/v sync times are accurate. For me anything over 2 frames is a no-go.

HDMI capture may work for Dolby Digital audio. I am not sure I know of any consumer or pro capture card that supports Dolby TrueHD or DTS. In fact, I am not sure what kind of capture card that you used to capture AC3. I know of only the Hauppauge Colossus that does that short of using something like a Teranex processor. How exactly did you match encoded AC3 audio to a video frame? It would be interesting to read more about your methodology and test results. Have you posted this on any forum/website?

I work on AV code for a living so I'm using my own capture software. It captures both AC3 and DTS from HDMI using a BlackMagic Intensity Pro PCIE card (non 4K version). You need to use their native SDK and some custom code to handle compressed audio. I'm not sure if any publicly available capture apps do this. Depending on the player you're capturing, you may also need to spoof the EDID because the MBIP only reports stereo PCM as supported and some devices will not send out AC3/DTS when they see that. Using my software, I get timestamps on each captured video and audio frame and can compare those to the presentation timestamps stored inside the original file being played. You need to understand the MPEG file formats and headers to do this manually. It's a bit tedious and time consuming to do this so it's been a while since I've done this for a recent Kodi build. It's possible things have regressed since version 14 I tested.