Tagging m4b is changing the timebase

The milliseconds of chapter times are changing when tagging m4b files. I’m not sure if it is isolated to m4b because I haven’t tested mp3 yet. This was reported here https://github.com/advplyr/audiobookshelf/issues/1060

When trying to track down why this is happening I noticed that m4b audiobooks I have that are using a 1/1000 timebase are being converted to 1/44100

I added a sample file named TimebaseSample.m4b here: https://github.com/advplyr/node-tone/tree/master/samples

Before tagging:

{
    "id": 1,
    "time_base": "1/1000",
    "start": 14043,
    "start_time": "14.043000",
    "end": 393113,
    "end_time": "393.113000",
    "tags": {
        "title": "Chapter 1"
    }
}

After tagging:

{
    "id": 1,
    "time_base": "1/44100",
    "start": 619296,
    "start_time": "14.042993",
    "end": 17336283,
    "end_time": "393.112993",
    "tags": {
        "title": "Chapter 1"
    }
}

The time base doesn’t really matter in the end but I think it may be related to the milliseconds getting changed.

[sandreas]

Ok, atldotnet released a fix for this today. I try to integrate this in the next days…

[sandreas]

I don’t think it has to do with ffmpeg since I’m only using ffprobe to read the data.

No this is definitely a tone issue. Thank you for your patience.

Here is how to reproduce with the sample file

Great thanks, I’ll have a look asap.

I don’t know enough about audio encoding to say if it matters to use 1/1000. The timebase is usually based on the sample rate so this may allow for more accuracy?

Neither do I. I have to do some research, what the problem is here but this is indeed an interesting problem, since I found NO hint that tone does something to change the timebase of the chapters. The 44100 just came to my mind because I also got a sample file for my tests to ensure, that values different from 1/1000 also do work as expected.

Now that I can reproduce the issue (hopefully), I’m sure I find out pretty quick, what the problem is.

One issue with changing the timebase seems to be it is changing the chapter start/end times.

That should never happen… it is definitely a problem, even if it is only a rounding issue.