Welcome to Doom9's Forum, THE in-place to be for everyone interested in DVD conversion. Before you start posting please read the forum rules. By posting to this forum you agree to abide by the rules. |
7th September 2011, 00:56 | #422 | Link |
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
The global "--norm" option basically is a shortcut for inserting the "gain -n" filter. Also the "norm" filter is nothing but an alias for "gain −n".
And yes, I tried it, just to be sure. But same effect
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ Last edited by LoRd_MuldeR; 7th September 2011 at 01:06. |
17th September 2011, 20:51 | #423 | Link | |
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
LameXP v4.03 Beta-2:
http://sourceforge.net/projects/lame...29/2011-09-17/ Quote:
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ Last edited by LoRd_MuldeR; 17th September 2011 at 23:31. |
|
18th September 2011, 17:56 | #424 | Link |
Registered User
Join Date: Aug 2008
Location: The Land Of Dracula (Romania - EU)
Posts: 934
|
here is the source code under GNU General Public License for a very fast normalizer...it has support only for 8/16bit simple header wavs up to 4gb...
_
__________________
if you ask a question and somebody give you the correct answer don't forget to leave a "thank you" note... Visit The Land Of Dracula (Romania - EU)! |
19th September 2011, 21:30 | #425 | Link |
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
LameXP running on Windows 8:
http://img841.imageshack.us/img841/2341/lamexpwin8.jpg Unfortunately it seems that some of the tools, including MediaInfo, will crash right away on Windows 8. Bummer! [EDIT] Seems like only the 64-Bit tools will crash. And yes, I'm using the x64 edition of Win8. [/EDIT] [EDIT2] It appears that MPRESS causes the issue. The original 64-Bit binary works fine, while the MPRESS-compressed one crashes right away. [/EDIT2] [EDIT3] As a temporary workaround you can use "--force-cpu-no-64bit" with the latest build in order to make LameXP work on Win8 64-Bit. [/EDIT3]
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ Last edited by LoRd_MuldeR; 20th September 2011 at 00:00. |
21st September 2011, 11:10 | #427 | Link | |
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
Quote:
Use a simple script, such as: FFAudioSource("C:\Some Path\Input.mkv") Possible, maybe. We would need to integrate a tool that can split the video file and "extract" the audio stream (e.g. MKVExtract for MKV files) for further processing. But this additional "splitting" step it doesn't fit into the current design of the software. So this would require bigger changes in the code...
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ |
|
29th September 2011, 22:13 | #428 | Link | |
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
LameXP v4.03 Beta-3:
Quote:
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ Last edited by LoRd_MuldeR; 16th October 2011 at 19:01. |
|
4th October 2011, 01:07 | #429 | Link |
Registered User
Join Date: Sep 2003
Location: Berlin, Germany
Posts: 3,079
|
Normalizing with SoX
Just noticed an unnecessary annoyance with the SoX "-ne" parameter. Whenever this parameter is used, SoX switches to 2-pass mode which of course doubles the filtering time.
While this is probably unavoidable for multichannel sources, it is complete nonsense to do 2 passes for stereo sources. SoX by itself is not smart enough to switch from "-ne" to just "-n" and does 2 passes on stereo sources which is a big waste of time. Could you add some code to LameXP to use "-ne" conditionally only if the source has more than 2 channels? Cheers manolito |
4th October 2011, 01:17 | #430 | Link |
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
Well, all the reason why we switched from "-n" to "-ne" is because "-n" is buggy. So, unfortunately, I think we can not switch back (now).
(Also, to my understanding, normalization always requires two passes. That's because the maximum peak in the file needs to be determined first, before the actual processing can start. Normalizing only within a local window would effectively "compress" the dynamics. IMHO that's not "normalization" any more)
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ Last edited by LoRd_MuldeR; 4th October 2011 at 13:16. |
4th October 2011, 14:35 | #431 | Link | ||
Registered User
Join Date: Sep 2003
Location: Berlin, Germany
Posts: 3,079
|
Quote:
Quote:
How else would you explain the performance hit when using "-ne" compared to just "-n" for stereo sources? I just did a couple of tests, and the results are reproduceable. Code:
Source: 2-ch WAV 1.4 GB Convert 2-ch WAV to normalized 2-ch WAV Commandline: SoX.exe infile outfile gain -n(e) -0.50 Execution time for "-ne": 16min 20sec Execution time for "-n": 12min 40sec So I would still lobby for using "-ne" only for multichannel sources... Cheers manolito |
||
4th October 2011, 14:46 | #432 | Link | ||
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
Quote:
Quote:
The latter will normalize all samples in all channels by the same factor, thus it is limited by the loudest sample in the loudest channel. At the same time the former will normalize all samples in a channel by the the same factor, but may choose a different factor for each channel. Consequently it is expected to get different results with "-ne", iff the peak values differ between your channels...
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ Last edited by LoRd_MuldeR; 4th October 2011 at 16:21. |
||
4th October 2011, 16:39 | #433 | Link | ||
Registered User
Join Date: Sep 2003
Location: Berlin, Germany
Posts: 3,079
|
Quote:
Quote:
Maybe the "-nb" option would be more suitable for stereo sources because it uses RMS values instead of peaks. This will result in a file with an even stereo balance (both channels have the same RMS value). This is often quite desirable, but probably not always... Cheers manolito |
||
4th October 2011, 16:48 | #434 | Link | |
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
Quote:
And I think, except for very rare cases, the peak levels of the channels of a Stereo recording should be identical (and should be equalized, if they differ).
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ |
|
4th October 2011, 17:27 | #435 | Link |
Registered User
Join Date: Aug 2008
Location: The Land Of Dracula (Romania - EU)
Posts: 934
|
no they are not...from my experience it is always a 0.5 to 1 db between left and right channels...
for 5.1 is even worse: center vs left+right front is 3-4 db and center vs left+right side/back is 4-8db... _
__________________
if you ask a question and somebody give you the correct answer don't forget to leave a "thank you" note... Visit The Land Of Dracula (Romania - EU)! |
4th October 2011, 17:35 | #436 | Link | |
Registered User
Join Date: Sep 2003
Location: Berlin, Germany
Posts: 3,079
|
Quote:
I have a background as a recording engineer (many years ago in the good old analog times), but if you do not believe me, test it for yourself: Take some of your favorite audio records (good quality please, maybe jazz), load them into WaveLab and do a "Global Analysis". This will give you the peak levels plus the RMS levels for each track (separately for each channel). Cheers manolito |
|
4th October 2011, 18:15 | #437 | Link |
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
Yes, I know that "sample value" and "perceived loudness" aren't the same at all and that RMS is a better loudness indicator. And I would never suggest to use the sample value to judge the loudness of a track. But we are not talking about average values here. Not even about average values for small segments. It's all about peak (maximum) values for complete tracks. Sure there are "high peaks which do not contribute to the perceived loudness of the track". And the number of such peaks may be very different between the channels. But each track/channel contains millions of individual sample values. So usually there is at least one (and one is sufficient for maximum calculation) of these "high" peaks in each channel. Actually my finding is that the maximum sample value generally is identical (or pretty close) between both channels of a Stereo recording, while both, the average and especially the maximum, RMS can differ significantly between the channels...
Example: http://img834.imageshack.us/img834/3...xpeakvalue.png
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ Last edited by LoRd_MuldeR; 4th October 2011 at 19:13. |
4th October 2011, 19:26 | #438 | Link | ||
Registered User
Join Date: Aug 2008
Location: The Land Of Dracula (Romania - EU)
Posts: 934
|
its time for some examples...
stereo (normalized after downmixing from 5.1) Quote:
original 5.1 Quote:
-ne for stereo will do: 0.00 -1.18 > 0.00 0.00 > meaning the right channel will be amplified by 1.18bd 0.00 -0.56 > 0.00 0.00 > meaning the right channel will be amplified by 0.56bd -0.43 0.00 > 0.00 0.00 > meaning the left channel will be amplified by 0.43bd 0.00 -0.26 > 0.00 0.00 > meaning the right channel will be amplified by 0.26bd -ne for original 5.1 will do: -0.66 -0.36 0.00 -3.15 -1.39 -2.35 > 0.00 0.00 0.00 0.00 0.00 0.00 > meaning the music (left/right channels) and the surround effect (side left/right channels) will be amplified a lot! -4.48 -3.18 0.00 -29.43 -19.77 -18.99 > 0.00 0.00 0.00 0.00 0.00 0.00 > meaning the music (left/right channels) and the surround effect (side left/right channels) will be amplified a lot! -6.42 -6.17 0.00 -12.28 -8.55 -8.75 > 0.00 0.00 0.00 0.00 0.00 0.00 > meaning the music (left/right channels) and the surround effect (side left/right channels) will be amplified a lot! -2.66 -2.70 0.00 -12.26 -10.03 -11.43 > 0.00 0.00 0.00 0.00 0.00 0.00 > meaning the music (left/right channels) and the surround effect (side left/right channels) will be amplified a lot! _
__________________
if you ask a question and somebody give you the correct answer don't forget to leave a "thank you" note... Visit The Land Of Dracula (Romania - EU)! Last edited by b66pak; 4th October 2011 at 19:53. |
||
6th October 2011, 23:31 | #439 | Link |
Software Developer
Join Date: Jun 2005
Location: Last House on Slunk Street
Posts: 13,248
|
After all I have decided to make the "channel equalization mode" an option.
Now you can even select "-n" again, but it will still fail with (some) multi-channel files, of course. The new build is available via auto-update...
__________________
Go to https://standforukraine.com/ to find legitimate Ukrainian Charities 🇺🇦✊ |
7th October 2011, 00:31 | #440 | Link |
Registered User
Join Date: Sep 2003
Location: Berlin, Germany
Posts: 3,079
|
Thanks very much...
Could you elaborate on how the three different normalization modes (max level, max energy, none) translate to SoX parameters? Couldn't find anything in the documentation... Cheers manolito |
Tags |
aac, aotuv, flac, lame, lamexp, mp3, mp4, ogg, oggenc, opus, vorbis |
Thread Tools | Search this Thread |
Display Modes | |
|
|