BUG- Auto Split- Inaccurate, Unreliable results when splitting at silence

Wavelab 9.5.50, Windows 10

Dear Steinberg team,
The auto-split function is not producing expected results when setting it to split audio at silence. It is a hit and miss.

I have a stereo wav consisting of drum samples, which are separated by absolute silence, edited in Cubase. I set the silence to be considered as below -140db. The resulting wav files contain various amounts of silence at the beginning of samples, making this feature useless, unfortunately.
I was so excited to use it, as it meant serious time savings, only to discover the results so unreliable, that it can not be used.

Please fix it and make it 100% reliable.

Thank you!

I set the silence to be considered as below -140db

Try another setting, such as -96 dB.

I tried. Still no good. Too many errors. I even ran the split function for the 2nd time on the resulting audio files.
This has to be sample-accurate and it is not even close. I am happy to provide a wav file that could be used to test it.

Any sort of silence-detecting function in Wavelab is terribly inaccurate and unreliable. I tried batch-removing silence at heads of those resulting files and that produced unsatisfactory results. I tried both “auto-detect silence” and setting the silence threshold to -14db RMS. The auto-detect did a poor job at detecting silences and selecting the threshold resulted in most files still having a bit of silence at heads of the files, which some had initial waveforms chopped off. No reliable sure way to do this. Very disappointing.
This is a serious shortcoming that should be addressed asap.

Where starts silence is something a bit subjective. Sample accuracy has no meaning, unless the audio contains digital silence. Since is uncommon, this particular case is not treated differently than the common case.

In my case, the silence is not subjective, but absolute. The audio contains absolute digital silence, amplitude nil. There is no excuse for Wavelab not to recognize it properly. At the very least, the software should be able to recognize absolute silence with absolute precision, but that is not the case. Not even close.

I can verify that -140 and -90 aren’t detecting digital silence/black/zero. I have to go to about -60 db to get any results for absolute silence, which doesn’t seem right. Also just trying it, I always get added silence at the end of clips or files (about 130 ms) when specifying “split at silences”.

I don’t get that when I use “cut head and tail” (which is what I usually use, to automatically trim clips. Again using a -60 db setting and adjusting afterward).

Goran have you tried a function like Autosplit in another program, and does it work better for you? I’ve never seen anything like it before.

I’ve gotten used to using it, and love it. But I agree, -140 and -90 should work to detect digital silence.

BTW what kind of audio process in which application generates an audio file with digital silence segments?

@PG I produce and edit drum samples in Cubase. I edit them perfectly, making starting points of evens at the very beginnings of waveforms. Events have spaces in between them with no audio, resulting in absolute silence when rendered. I was trying to export sections of audio consisting of multiple drum samples (separated by silence), then use Wavelab to automatically and accurately split them into individual drum samples and name them accordingly. Since these are drum samples, sample-accurate beginnings are crucial.

@bob99 Thank you. I have not yet found a program that does this well. I tried one, but it was also inaccurate. You would think that this would be a piece of cake for a software to complete.

I also remember having trouble with this. I often edit and mixdown multichannel files in Reaper and need to eliminate parts with digital silence after import in WL. Hope this gets fixed, at least in V10.