Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: encoding voice... bad sound (Read 9594 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

encoding voice... bad sound

Hi everyone. This is my first post here and I hope you find it useable or interesting. I am no expert or anything when it comes to lossy audio compression. I just know what I think sound good and/or bad.

I would like to show you what I discovered when I used full one band stereo widening in Samplitude on audio with voice encoded with the latest OggdropXp with AoTuVb3 optimizing.

The ogg sample was encoded (q5) and then decoded with the OggdropXp version described above, and then processed in Samplitude using the built in stereo enhancer and saved as wav to not degrade the audio any further.

Uncompressed (unprocessed)
Uncompressed (processed. Full one band stereo widening)
OggOgg (processed. Full one band stereo widening)

I also encoded the audio in AAC (-b 160 and no low pass) using FAAC which didn't give me any audible artifacts.
What is your thoughts on this? Is this something that should be fixed quickly?

I have noticed that using stereo enhancers, you can easily hear if lossy compression has added artifacts. I always use this method to hear how good codecs are. It probably isn't the best and only way to hear/detect artifacts.

Best Regards
Rickard Gerthsson
http://www.damnsundaydrivers.com

encoding voice... bad sound

Reply #2 –
I do that too, I use soundforge function called "pan\expand"
I also can do it on my headphones, I just switch it to 3d setting
This happens as the temporal masking affect and ear (nspsytune) models

They are desinged in codecs to sound good to human ear but not desingned to 3d sounding and stereo widening etc.. that's why you hear artifacts because lame and other codecs encode to human ear listining capabilities but not to audio processing

If you did the audio processing to the source then encoded it it would sound much better

EDIT: Typo

encoding voice... bad sound

Reply #3 –
Quote
This happens as the temporal masking affect and ear (nspsytune) models
[a href="index.php?act=findpost&pid=275506"][{POST_SNAPBACK}][/a]

So this is something which is intended to happen becuase we can't hear it if we do not apply effects to the lossy audio? So this might show why lossless is best for editing and lossy is best for.... hm... listen to audio which are intended to be listened to "as it is"?

Aac didn't have this problem, but does it might have other "bad" things which ogg doesn't have?

Best Regards
Rickard Gerthsson

Edit: Oh, I now saw that Corezode answered my question about using lossy audio to listen to as it is and not for future archiving(?). Thanks for the help guys!

encoding voice... bad sound

Reply #4 –
hey punkrockdude

some stereo widening program (spatial) IS BAD (e.g.) I do have a "spatial" DSP in my windows playback control which creative driver includes....  it sounds terrible when i activate it!!!

anyway what stereo widening DSP are you using

I find foobar2000's spatial 3D quite good.....you can give it a try
it offers a few settings

Quote
The ogg sample was encoded (q5)


oh I missed this point...NOTE...ogg use LOSSY stereo at -q 5, this might the reason there is artifact

FAAC uses LOSSLESS stereo!!

try (-q 6), see whether it solves this problem

encoding voice... bad sound

Reply #5 –
Quote
hey punkrockdude

some stereo widening program (spatial) IS BAD (e.g.) I do have a "spatial" DSP in my windows playback control which creative driver includes....   it sounds terrible when i activate it!!!

anyway what stereo widening DSP are you using

I find foobar2000's spatial 3D quite good.....you can give it a try
it offers a few settings

Quote
The ogg sample was encoded (q5)


oh I missed this point...NOTE...ogg use LOSSY stereo at -q 5, this might the reason there is artifact

FAAC uses LOSSLESS stereo!!

try (-q 6), see whether it solves this problem
[a href="index.php?act=findpost&pid=275628"][{POST_SNAPBACK}][/a]


Hi Korttim.

I used the built in stereo enhancer in the sequencer Samplitude. I also tried encoding it in -q6 and -q7, though it sounds better the higher value I use, I still shows quite much artifatacts even at -q7

Regards
Rickard Gerthsson

encoding voice... bad sound

Reply #6 –
Quote
still shows quite much artifatacts even at -q7


what artifact? hisshing?

i think it will be quite useful to developers........

encoding voice... bad sound

Reply #7 –
Quote
Quote
still shows quite much artifatacts even at -q7


what artifact? hisshing?

i think it will be quite useful to developers........
[a href="index.php?act=findpost&pid=275640"][{POST_SNAPBACK}][/a]


Even at -q9 i can hear the artifacts using stereo widening.
Man, I am so lost. Should I use AAC or Vorbis? I have tried aac and I ind it better sounding, but I don't want to let go of vorbis right away. If these artifacts wasn't there, I would not at all consider going over to AAC.

encoding voice... bad sound

Reply #8 –
Quote
Hi everyone. This is my first post here and I hope you find it useable or interesting. I am no expert or anything when it comes to lossy audio compression. I just know what I think sound good and/or bad.

I would like to show you what I discovered when I used full one band stereo widening in Samplitude on audio with voice encoded with the latest OggdropXp with AoTuVb3 optimizing.

The ogg sample was encoded (q5) and then decoded with the OggdropXp version described above, and then processed in Samplitude using the built in stereo enhancer and saved as wav to not degrade the audio any further.

Uncompressed (unprocessed)
Uncompressed (processed. Full one band stereo widening)
OggOgg (processed. Full one band stereo widening)

I also encoded the audio in AAC (-b 160 and no low pass) using FAAC which didn't give me any audible artifacts.
What is your thoughts on this? Is this something that should be fixed quickly?

I have noticed that using stereo enhancers, you can easily hear if lossy compression has added artifacts. I always use this method to hear how good codecs are. It probably isn't the best and only way to hear/detect artifacts.

Best Regards
Rickard Gerthsson
http://www.damnsundaydrivers.com
[a href="index.php?act=findpost&pid=275501"][{POST_SNAPBACK}][/a]



Please use the encoder of reference by my web page. I do not hear a large problem like the above-mentioned sample.
 

encoding voice... bad sound

Reply #9 –
I tried your reference encoder and it sounds the same. I want to say once again that these artifacts are only heard when I apply extreme stereo widening to the already encoded file.

encoding voice... bad sound

Reply #10 –
Quote
NOTE...ogg use LOSSY stereo at -q 5[a href="index.php?act=findpost&pid=275628"][{POST_SNAPBACK}][/a]

Not true! Not really.

Oggenc 1.0 and higher use LOSSLESS from q6, but q5.99 and below can be lossless too. Lossy channel coupling starts being used from 5.99, but lossless channel coupling can still be used if necessary.

encoding voice... bad sound

Reply #11 –
Quote
I want to say once again that these artifacts are only heard when I apply extreme stereo widening to the already encoded file.
[{POST_SNAPBACK}][/a]

[a href="http://www.catb.org/~esr/jargon/html/D/Don-t-do-that-then-.html]http://www.catb.org/~esr/jargon/html/D/Don...that-then-.html[/url]

encoding voice... bad sound

Reply #12 –
Quote
Quote
I want to say once again that these artifacts are only heard when I apply extreme stereo widening to the already encoded file.
[{POST_SNAPBACK}][/a]

[a href="http://www.catb.org/~esr/jargon/html/D/Don-t-do-that-then-.html]http://www.catb.org/~esr/jargon/html/D/Don...that-then-.html[/url]
[a href="index.php?act=findpost&pid=277507"][{POST_SNAPBACK}][/a]

encoding voice... bad sound

Reply #13 –
Quote
I tried your reference encoder and it sounds the same. I want to say once again that these artifacts are only heard when I apply extreme stereo widening to the already encoded file.
[a href="index.php?act=findpost&pid=277505"][{POST_SNAPBACK}][/a]


What I tried is the "uncompressed_stereo_widening.wav" sample shown you. Then, especially I did not encounter a problem.
It does not bring a severe result like "ogg_stereo_widening.wav."

encoding voice... bad sound

Reply #14 –
Quote
Quote
I tried your reference encoder and it sounds the same. I want to say once again that these artifacts are only heard when I apply extreme stereo widening to the already encoded file.
[a href="index.php?act=findpost&pid=277505"][{POST_SNAPBACK}][/a]


What I tried is the "uncompressed_stereo_widening.wav" sample shown you. Then, especially I did not encounter a problem.
It does not bring a severe result like "ogg_stereo_widening.wav."
[a href="index.php?act=findpost&pid=277512"][{POST_SNAPBACK}][/a]


Try the "Uncompressed (unprocessed)" and encode it to ogg. Then open the ogg with a program which can do stereo widening and listen.

encoding voice... bad sound

Reply #15 –
Quote
Quote
Quote
I want to say once again that these artifacts are only heard when I apply extreme stereo widening to the already encoded file.
[{POST_SNAPBACK}][/a]

[a href="http://www.catb.org/~esr/jargon/html/D/Don-t-do-that-then-.html]http://www.catb.org/~esr/jargon/html/D/Don...that-then-.html[/url]
[a href="index.php?act=findpost&pid=277507"][{POST_SNAPBACK}][/a]


[a href="index.php?act=findpost&pid=277511"][{POST_SNAPBACK}][/a]

Seriously, though.  This is lossy perceptual encoding; when you alter the parameters of the perception (by heavy digital postprocessing) then all bets are off.  If this problem isn't apparent in another format then that's only due to said format not being clever enough to throw away differences that would be inperceptable without such postprocessing, and it's not up to an encoder/format (except perhaps as hacky situation-specific encoding flags) to anticipate random combinations of postprocessing a user might want to apply.

Perhaps consider putting your speakers further apart.   

encoding voice... bad sound

Reply #16 –
Quote
Quote
Quote
Quote
I want to say once again that these artifacts are only heard when I apply extreme stereo widening to the already encoded file.
[{POST_SNAPBACK}][/a]

[a href="http://www.catb.org/~esr/jargon/html/D/Don-t-do-that-then-.html]http://www.catb.org/~esr/jargon/html/D/Don...that-then-.html[/url]
[a href="index.php?act=findpost&pid=277507"][{POST_SNAPBACK}][/a]


[a href="index.php?act=findpost&pid=277511"][{POST_SNAPBACK}][/a]

Seriously, though.  This is lossy perceptual encoding; when you alter the parameters of the perception (by heavy digital postprocessing) then all bets are off.  If this problem isn't apparent in another format then that's only due to said format not being clever enough to throw away differences that would be inperceptable without such postprocessing, and it's not up to an encoder/format (except perhaps as hacky situation-specific encoding flags) to anticipate random combinations of postprocessing a user might want to apply.

Perhaps consider putting your speakers further apart.   
[a href="index.php?act=findpost&pid=277518"][{POST_SNAPBACK}][/a]


Ok, thanks. I just wanted to show you my "discovery" and I thought that it might be useful. Now I understand that lossy formats are made to be "listened to as it is". No altering (stereo widening, eq:ing, etc), for that I'll use FLAC.

Regards
Rickard Gerthsson
Sweden

encoding voice... bad sound

Reply #17 –
Quote
Quote
NOTE...ogg use LOSSY stereo at -q 5[a href="index.php?act=findpost&pid=275628"][{POST_SNAPBACK}][/a]

Not true! Not really.

Oggenc 1.0 and higher use LOSSLESS from q6, but q5.99 and below can be lossless too. Lossy channel coupling starts being used from 5.99, but lossless channel coupling can still be used if necessary.
[a href="index.php?act=findpost&pid=277506"][{POST_SNAPBACK}][/a]



I remember oggenc does have loseless stereo build in any -q but it was missing since version 1. I still miss that features until today.

Could someone please put into oggenc wish list for the options??
I'm not coding guru, I heard the options can be tweak but need re-compile.