I have recently changed to using 24bit audio in audacity. I did some listening tests and found there was a very big difference, even though my initial recordings were in 16bit from my usb mixer (Behringer 1204) the resulting mix of multiple recording, stored in Audacity at 32bit float, when mixed produced a higher definition audio, in the same way that a panoramic picture made with multiple original jpg comes out at a much higher definition in the panorama. Same principal applies to audio.

So I have standardised all my music production (albums and music videos) at 24bit 48000 FLAC. These are master recording which are held for the future benefit of mankind. I am using Audacity for audio and Kdenlive for video production. Thats all I use. On Linux Mint 17.

Just recently I have been having problems rendering files with many small audio flac files mixed with a little video, and rendered into my usual Ogg Theora format - which I had identified as the best long term archival format. The videos produced would sometimes not play in Gnome Video, yet allways in VLC. This was a warning sign that there was some problem with some of these files.

I managed to find a workaround that always works - this is to change the container to Matroska - something I didnt want to do at first, yet having searched the topic it seems there is some move towards this matroska away from ogg.

Whilst examining this issue I discovered that my flac files, which I thought were being encoded without change, were ending up, after rendering, at 16bit rather than 24bit. This means that I am losing quality in my masters. Is there any way to force Kdenlive to output at 24bit? I tried sample_fmt=s24 yet that didn't seem to do anything.

I am posting here to make the community aware of what people like me are using the software for, and their concerns. My work is similar to that done by Alan Lomax and involves preservation of our Blues Music Heritage. Whatever comes out of Kdenlive ends up as the master and the original unedited raw videos are scrapped (too complex to document). The music is overdubbed and combined with the video footage at high resolution. There is no high resolution audio on the originasl video footage.

If anyone can pass comment on the direction I am moving in, vs the general community, and any solutions to the 16bit 24bit flac issue. Since Neil Young established the Flac format as the universal format of High Quality Audio (since the better DSD format is proprietary at this point in time) I would expect people to be moving in that direction, like myself. Any comments on that, please?

Thanks and best wishes to the group

The 16 bit vs 24 bit argument is ****. Human ears can not hear the difference - when you are comparing the same dynamic range of audio. See https://www.youtube.com/watch?v=6Zwjn7hgFV4

I tried sample_fmt=s24 yet that didn't seem to do anything.

Only indirectly (as far as I know):

Code: Select all

properties=WAV ar=48000 acodec=pcm_s24le

All available:

Code: Select all

melt -query "audio_codecs"

ttguy wrote:Human ears can not hear the difference

Has nothing to do with it, see https://en.wikipedia.org/wiki/Audio_bit_depth

Stetsbequem wrote:

ttguy wrote:Human ears can not hear the difference

Has nothing to do with it, see https://en.wikipedia.org/wiki/Audio_bit_depth

Are you saying this has nothing to do with the whole Pono scam from Neil Young?
http://www.theguardian.com/technology/2 ... khz-review

Or are you saying that you don't care if the human ear can not hear the difference?

I'm sorry, my English is not enough to answer you in detail. I also work as long as the processing is completed in 24-bit (recording)/32-bit (DAW intern). And of course, the project is saved for the future, for subsequent revisions (remastering). All technical reasons, that was my concern.

Finally, after the mastering and for the ultimate consumer, the whole will be converted to 16-bits per export. But if one, like Mr. Young, can hear or feel the differences, why not 24-bit/44..192kHz?

Stetsbequem wrote:the project is saved for the future, for subsequent revisions (remastering). All technical reasons, that was my concern.

Fair enough.

I was wondering if I can export 32-bit float WAV with my videos to be able to measure the loudness (ebur128) and apply gain and limiting to normalize my videos to say -18 LUFS.

I'll try if I can do that with the mentioned tricks. I hope MLT will not truncate my bit depth in the process somewhere.

unfa,
If Jack is running, the Ebumeter can be used for this. The data does not have to be exported. Of course, it runs in real time, but you get a good impression.

If you can export more than 24-bit, please post the profile.