Sound extraction without using FFMPEG


Ad detector
Is it possible to access the audio component of a .ts file without using the ffmpeg program?

The reason for asking is that detectads has been using the frame numbers in the output from ffmpeg to create the bookmark time stamps and nicesplice cut points by frame number.

This has proved very successful with good quality SD recording and most HD recordings but recent testing on HD recordings has shown problems with parallel recordings and low quality recordings .

I have discovered that for HD files that the nts file frame rate increases to 26-27fps when multiple recordings are being made meaning the calculated cut points become increasingly inaccurate though calculated times are still OK provided the recording quality is OK.

I have fairly low signal strength and whilst recording SD is OK I often suffer dropouts on HD so I have one recording which according to the Humax is 61 minutes and about 95K frames but when processed by ffmpeg it was reported as 58 minutes and 87K frames - with such discrepancies there is no chance of getting the ad breaks right.

So what I need is way of directly correlating the silences in the audio stream directly with the Humax frame numbers / time stamps without the reframing by ffmpeg.

Are there any examples of code that extract the audio directly from a ts file or clear explanations of the file structure
It's unfinished, but stripts -S <file> demultiplexes the raw audio stream into a corresponding .audio file.
It's unfinished, but stripts -S <file> demultiplexes the raw audio stream into a corresponding .audio file.
Out of interest, are there any problems with the LATM headers? The AAC audio itself is fine, but it is surprising how little support there is for LATM. Even with ffmpeg, support is deemed 'experimental'. I have managed to use ffmpeg to convert H.264/ LATM-AAC ts files to mp4 on the HDR-FOX, but it is slow (a 30 minute recording takes about 1 hour 10 minutes) and error prone.