Using CC2019 I am converting a large number of accurately timed wav files to mp3. I want the mp3 files to be the same duration as the wav's but exporting to 192kbps CBR at 44100/16 stereo, a 59:00 long mp3 is about 6 seconds shorter in run time, (i.e. 58:54), a 2:59:00 wav to mp3 exports as 2:58:44. I am having to use Stetch and Pitch with special timings in order to get the required duration.
Fortunately the discrepancy is consistent, so for example, batching a number of 59.00 minute files with an S&P preset set at a new duration of 59.06 then exporting to mp3 seems to do the trick, but why is it happening? It is a PITA to have to experiment to find a setting that works for different durations. The conversions also take time, the 59 min progs take about 25 mins.
The win10 pc is reasonably high end, it will batch 15 wavs simultaneously but I get the same issues on a win10 box running CS6 which will only handle 3 simultaneously.
Copy link to clipboard
We get complaints about this on a regular basis. Clearly you haven't fallen into the trap of using VBR, which is pretty much guaranteed to give you false results, but even using CBR the sums don't always work out. How are you establishing that the actual timing errors exist? Is this just based on the reported durations, or do they actually run for an incorrect duration?
Last I heard, Adobe were still buying in an MP3 package, so its behaviour isn't anything that Adobe can do anything about directly. We've had reports that different versions (with different encoder versions) don't all behave the same, as well. It might be worth (even if you only do it as a comparison) encoding one of these files with another encoder - LAME for instance - and seeing if you get the same results.
The other thing to note is that if an hour's worth of material comes out a real six seconds short, that represents a pitch shift up of 0.016% - which I would have thought would be audible if you played the decoded mp3 file against the original wav, appearing as a significant phase error, getting worse as it progresses. It may well be worth checking decodes from a few different encoders to see if these are real differences, or just reported errors. The MP3 decode standard says that they all have to be decoded the same way, so if there are errors, it would definitely be in the encoding.
Obviously you can't select an MP3 encoder in Audition, and if you want to do any sort of comparisons my advice would be to use Exact Audio Copy - which is free, and always has been very good.
The timings are actual. Windows and our transmission automation software reports the durations correctly and the mp3 would play out short if not adjusted. Obviously this isn't good in an automation system where things are timed to the second and we keep a full operational backup using the mp3 versions to save space in a 16/7 transmission schedule. Without the stretch & pitch process they would be unusable. It's just a time consuming tedious process.
When I get a moment I will try Exact Audio and see how it compares.