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: lossyWAV Development (Read 571429 times) previous topic - next topic
0 Members and 2 Guests are viewing this topic.

lossyWAV Development

Reply #1125 –
Quote from: Nick.C on
Quote from: halb27 on
Quote from: Nick.C on
... quality_signal_to_noise_ratios    : array[0..Quality_Presets] of Double  = (18,18.87,19.81,20.8,21.86,23,24.21,25.51,26.91,28.4,30); ...
As this makes things more defensive: go ahead.
But why these strange steps like 18.87?
I was looking for a smooth curve, so I worked out the power required to make 18 translate to 30 in 10 steps (i.e. snr:=power(snr[i-1],z)).

I see. And it's all internal anyway.

Quote from: GeSomeone on
... I'd like to ask Halb27 if he's willing to do an ABX of (current) -q 5 vs.  -q 6 for those 4 problem samples. ...

OK. Luckily it's just 3 samples cause MartenotWaves was alright with -q 5 (triangle as well in the sense that I couldn't abx it, but there is a suspicion that it isn't perfect as I started with 4/4).

I could not abx badvilbel and triangle -q 5 vs. -q 6.
My result for Under the Boardwalk was 7/10 (the same as -q 5 vs. original).

Obviously this isn't a big issue for me with -q 5, but I'd like to play it safe when using lossyWAV.
Apart from that my 58 year old ears are a bit trained now to these samples, but there are certainly ears out there which perform a lot better.
Hopefully we get a lot of more listening experience feedback.
lame3995o -Q1.7 --lowpass 17

lossyWAV Development

Reply #1126 –
Using the proposed revision to the -snr parameter, the following bitrates were achieved when I processed my 53 problem sample set:

Code: [Select]
|-------------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|
|  lossyWAV   |  -q 0   |  -q 1   |  -q 2   |  -q 3   |  -q 4   |  -q 5   |  -q 6   |  -q 7   |  -q 8   |  -q 9   |  -q 10  |
|-------------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|
| beta v0.9.6 | 318kbps | 338kbps | 364kbps | 394kbps | 431kbps | 472kbps | 500kbps | 529kbps | 557kbps | 584kbps | 611kbps |
|-------------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|
| variant #1  | 327kbps | 346kbps | 370kbps | 400kbps | 435kbps | 475kbps | 502kbps | 530kbps | 557kbps | 584kbps | 611kbps |
|-------------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|
| variant #2  | 327kbps | 348kbps | 373kbps | 403kbps | 438kbps | 477kbps | 504kbps | 531kbps | 558kbps | 585kbps | 611kbps | 
|-------------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|---------|


However, looking at that, not enough is done around the -q 5 mark, so I'm going to try:

variant #2: quality_signal_to_noise_ratios : array[0..Quality_Presets] of Double = (18.0,19.2,20.4,21.6,22.8,24.0,25.2,26.4,27.6,28.8,30.0);

instead of

variant #1: quality_signal_to_noise_ratios : array[0..Quality_Presets] of Double = (18.0,18.9,19.8,20.8,21.9,23.0,24.2,25.5,26.9,28.4,30.0);

instead of

quality_signal_to_noise_ratios : array[0..Quality_Presets] of Double = (16.0,17.0,18.0,19.0,20.0,21.0,22.8,24.6,26.4,28.2,30.0);

I am a bit happier with the spread of the bitrate outputs from the various quality presets. I'll have a think and probably post beta v0.9.7 tomorrow.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1127 –
IMO the more defensive -snr values are welcome especially for the low bitrate settings. It's not so important for -q 5+ IMO.

I wonder about something else. Do we have a specific problem with impulses? (eig - a very serious mp3 pre-echo problem - shows the worst performance at -q 0, and it's so bad around the impulses, and Under the Boardwalk seems to have a small problem at -q 5, and the slightly changed pitch I perceive is with drums. I also remember that AlexB's very first lossyWAV -3 listening experience led to a changed pitch detection, and his sample is full of percussion.).
May be impulses should be taken special care of. Trying to improve is possible without hard listening tests by striving at a good eig performance at -q 0. Maybe a special impulses detection could help which automatically lowers the number of bits to remove drastically?
lame3995o -Q1.7 --lowpass 17

lossyWAV Development

Reply #1128 –
How to search for an impulse though.....?

Might one approach be to split the codec-block into 8/15 (16/31?) 50% overlapping chunks and take RMS values of the samples in each chunk, then look at the relative magnitudes of the per-chunk-RMS-results to try to spot an impulse?

Or, perform 16 (or 32) sample FFT's (8/15 or 4/7, 50% overlapping) and look at the maximum bin result in each?

Or, just use the maximum bin (skewed / unskewed?) result from the 9 x 64 sample FFT's already calculated per channel per codec-block and try to spot the high value?
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1129 –
Quote from: Nick.C on
How to search for an impulse though.....?

Might one approach be to split the codec-block into 8/15 (16/31?) 50% overlapping chunks and take RMS values of the samples in each chunk, then look at the relative magnitudes of the per-chunk-RMS-results to try to spot an impulse?

Or, perform 16 (or 32) sample FFT's (8/15 or 4/7, 50% overlapping) and look at the maximum bin result in each?

Or, just use the maximum bin (skewed / unskewed?) result from the 9 x 64 sample FFT's already calculated per channel per codec-block and try to spot the high value?

I have no idea what's best. All of your proposals make sense to me.
lame3995o -Q1.7 --lowpass 17

lossyWAV Development

Reply #1130 –
Quote from: halb27 on
Do we have a specific problem with impulses? (eig - a very serious mp3 pre-echo problem - shows the worst performance at -q 0, [..] Under the Boardwalk seems to have a small problem at -q 5, [..]).
Maybe impulses should be taken special care of.

Quote from: Nick.C on
How to search for an impulse though.....?

Just trying to think along in finding an approach for this (just a bunch of questions to consider, I'm afraid)
First of all: is this new or more severe than in previous versions? (if that's true .. what was changed).
Can the transients be catched with one of the existing mechanisms? e.g. Does it get better when raising -nts (I know it's hidden from the interface right now). The -nts value distribution (over the -q's) has been changed lately, is it working properly?
Less likely, does adding FFT's help?

Could it be -snr could help this too? (try with a high quality_signal_to_noise_ratio).

It is a suspicion from me too that sounds like drums with hi-hats sometime sound not as "crisp" at settings below -q 5. But you can't take my word for it as I'm terrible at ABX, after 2x I usually hear a no difference anymore.
In theory, there is no difference between theory and practice. In practice there is.

lossyWAV Development

Reply #1131 –
Quote from: Nick.C on
How to search for an impulse though.....?

I've looked at eig and Under The Boardwalk using a wav editor.
Maybe a very simple procedure does it: watch the difference of the value of two consecutive samples. If the absolute value of the difference is larger than a certain threshold: reduce the number of bits to remove depending on the size of the difference.
Make the threshold and number-of-bits-to-remove-depence on the sample difference more demanding for the higher -q settings.
Here's the critical beginning of eig in case you haven't got eig, Nick, if you want to play with it.
lame3995o -Q1.7 --lowpass 17

lossyWAV Development

Reply #1132 –
Quote from: halb27 on
Quote from: Nick.C on
How to search for an impulse though.....?
I've looked at eig and Under The Boardwalk using a wav editor.
Maybe a very simple procedure does it: watch the difference of the value of two consecutive samples. If the absolute value of the difference is larger than a certain threshold: reduce the number of bits to remove depending on the size of the difference.
Make the threshold and number-of-bits-to-remove-depence on the sample difference more demanding for the higher -q settings.
Here's the critical beginning of eig in case you haven't got eig, Nick, if you want to play with it.
Many thanks for the insight - I'll get coding to implement a "net" to find the maximum absolute difference between samples for each channel in a codec-block.

I have already implemented a search for the bin with the maximum value, the simple average value and the minimum value for each FFT analysis.

Thanks for something else to chew on!
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1133 –
Just a remark, Nick, as you like so much to use your 53 sample set:
For judging about the negative impact this impulse-defensive idea will have on bitrate: please use a set of regular music to get an impression of the consequences. With problem samples it's welcome that bitrate goes up, with regular music it's not. My just 12 entire tracks set of regular music is encoded quickly, and the bitrate results have always been close to your more advanced multi-album test. So I suggest you use just a selection of a couple of full length tracks. Just take care a bit that the musical content of the tracks selected isn't too similar.
lame3995o -Q1.7 --lowpass 17

lossyWAV Development

Reply #1134 –
One way to throw more bits at impulses is to use a shorter FFT, e.g. 32.

I'm not saying you should, only that it could be worth trying. It'll "see" the space around impulses more, which may be a good or bad thing overall.

Cheers,
David.

lossyWAV Development

Reply #1135 –
Quote from: 2Bdecided on
One way to throw more bits at impulses is to use a shorter FFT, e.g. 32.

I'm not saying you should, only that it could be worth trying. It'll "see" the space around impulses more, which may be a good or bad thing overall.

Cheers,
David.
Many thanks, David, for the advice - it was also the simplest by far to implement at the expense of additional process time.

lossyWAV beta v0.9.7 attached to post #1 in this thread.

[edit] Processed 53 sample problem set bitrates: (10 album test set to follow).

Code: [Select]
|----------------------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
|      lossyWAV        | -q 0  | -q 1  | -q 2  | -q 3  | -q 4  | -q 5  | -q 6  | -q 7  | -q 8  | -q 9  | -q 10 |
|----------------------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
| beta v0.9.6          |318kbps|338kbps|364kbps|394kbps|431kbps|472kbps|500kbps|529kbps|557kbps|584kbps|611kbps|
|----------------------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
| beta v0.9.7          |327kbps|346kbps|370kbps|400kbps|435kbps|475kbps|502kbps|530kbps|557kbps|584kbps|611kbps|
|----------------------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
| beta v0.9.7 -impulse |342kbps|360kbps|383kbps|412kbps|446kbps|485kbps|513kbps|540kbps|567kbps|594kbps|619kbps|
|----------------------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1136 –
Quote
E:\Utils\LossyWAV>lossyWAV.exe test.wav -analyses
%lossyWAV Error% : No analyses value given.

E:\Utils\LossyWAV>lossyWAV.exe test.wav -analyses 2
lossyWAV beta v0.9.7, Copyright © 2007,2008 Nick Currie.
lossyWAV is issued with NO WARRANTY WHATSOEVER and is free software.

%lossyWAV Error% : Incorrect option: "-analyses"


It seems something was broken in 0.9.6 -> 0.9.7 change

lossyWAV Development

Reply #1137 –
Quote from: lvqcl on
Quote
E:\Utils\LossyWAV>lossyWAV.exe test.wav -analyses
%lossyWAV Error% : No analyses value given.

E:\Utils\LossyWAV>lossyWAV.exe test.wav -analyses 2
lossyWAV beta v0.9.7, Copyright © 2007,2008 Nick Currie.
lossyWAV is issued with NO WARRANTY WHATSOEVER and is free software.

%lossyWAV Error% : Incorrect option: "-analyses"


It seems something was broken in 0.9.6 -> 0.9.7 change
Thanks for that, revised version v0.9.7 going up now....

Bitrates for 10 album test set:

beta v0.9.6 : -q 10: 573kbps; -q 5: 417kbps; -q 0: 286kbps
beta v0.9.7 : -q 10 -impulse: 580kbps; -q 5 -impulse: 429kbps; -q 0 -impulse: 310kbps
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1138 –
I've spent nearly an hour trying to use the batch file that's in the wiki for foobar, but it just won't go no matter what edits I make.
Could someone post a copy of a flossy bat that works? -- preferably using just a c: drive.
I did get the latest beta to work from the command line and my initial impression is very favorable -- "Yours Is No Disgrace" 29.0Mb vs. 65.2Mb is a worthwhile saving.
FWIW, there might be others like me who hadn't realized that this is a pre-processor for FLAC rather than a codec that creates altered WAV files. 
An obvious application for me would be converting language instruction CDs to greatly reduced semi-lossless.  There'd likely be time savings using this rather than downsampling to 32000 and converting to mono in CoolEdit before converting to FLAC.  An additional step has been amplifying 6db, but, perhaps there's a foobar plgin for that?
Anyway, thanks much for your efforts.  The original idea was a cunning and clever one, but there's obviously been plenty of perspiration since...

lossyWAV Development

Reply #1139 –
Quote from: ckjnigel on
I've spent nearly an hour trying to use the batch file that's in the wiki for foobar, but it just won't go no matter what edits I make.
Could someone post a copy of a flossy bat that works? -- preferably using just a c: drive.
I did get the latest beta to work from the command line and my initial impression is very favorable -- "Yours Is No Disgrace" 29.0Mb vs. 65.2Mb is a worthwhile saving.
FWIW, there might be others like me who hadn't realized that this is a pre-processor for FLAC rather than a codec that creates altered WAV files. 
An obvious application for me would be converting language instruction CDs to greatly reduced semi-lossless.  There'd likely be time savings using this rather than downsampling to 32000 and converting to mono in CoolEdit before converting to FLAC.  An additional step has been amplifying 6db, but, perhaps there's a foobar plgin for that?
Anyway, thanks much for your efforts.  The original idea was a cunning and clever one, but there's obviously been plenty of perspiration since...
Is the batch file on a path with spaces in it? I found this to be an elusive problem to solve. That is why my batch file is in a simple <drive>:\BIN\ directory, as are flac.exe and lossyWAV.exe - also, ensure that the batch file references the correct locations of the two relevant .exe files.

[edit2] Oh, and it's a lossy pre-processor which produces modified WAV files. It works with other codecs apart from FLAC, although I use FLAC by preference as it is compatible with TCPMP v0.81 on my iPAQ. [/edit2]

[edit] For example: flossy.bat
Code: [Select]
@echo off
c:\data_nic\bin\lossyWAV %1 %3 %4 %5 %6 %7 %8 %9 -low -nowarn -quiet
c:\data_nic\bin\flac.exe -5 -f -b 512 "%~N1.lossy.wav" -o"%~N2.flac" 
del "%~N1.lossy.wav"
with the batch file, lossyWAV.exe and flac.exe in the same directory, i.e. C:\DATA_NIC\BIN\ and called from foobar2000 with:

Encoder: cmd.exe

Extension: lossy.flac (NOT .lossy.flac!!!)

Parameters: /d /c c:\data_nic\bin\flossy.bat %s %d <insert your parameters here>

example parameters could be: -q 4 -impulse [/edit]
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1140 –
Thank you Nick, for your new version.

I'm too tired now for abxing higher quality settings, but gave it a try using -q 0 -impulse for eig.
Yes, there's an abxable improvement with eig.

Bitrate increase for regular music isn't very remarkable: my regular music track set went up from 417 kbps (v0.9.6 -q 5) to 427 kbps (v0.9.7 -q 5 -impulse).

Hope I can do more listening tests tomorrow.

What -spf values are using for the 32 samples FFTs, Nick?
lame3995o -Q1.7 --lowpass 17

lossyWAV Development

Reply #1141 –
Quote from: halb27 on
Thank you Nick, for your new version.

I'm too tired now for abxing higher quality settings, but gave it a try using -q 0 -impulse for eig.
Yes, there's an abxable improvement with eig.

Bitrate increase for regular music isn't very remarkable: my regular music track set went up from 417 kbps (v0.9.6 -q 5) to 427 kbps (v0.9.7 -q 5 -impulse).

Hope I can do more listening tests tomorrow.

What -spf values are using for the 32 samples FFTs, Nick?
I iterated a few times until I just used 22223 (the same as for the 64 sample FFT) as increasing the 2's results in a higher bitrate(!). 22222 was also higher in bitrate - 22223 seems to be a sweet spot (some averaging, but not too much).
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1142 –
Quote from: Nick.C on
I iterated a few times until I just used 22223 (the same as for the 64 sample FFT) as increasing the 2's results in a higher bitrate(!). 22222 was also higher in bitrate - 22223 seems to be a sweet spot (some averaging, but not too much).

Thank you, Nick.
Do you mind making -spf temporarily available to the user again? (I'm only interested in playing around with the -spf setting for the 32 samples FFT, I'm just curious about the quality of 22222).
lame3995o -Q1.7 --lowpass 17

lossyWAV Development

Reply #1143 –
Quote from: halb27 on
Quote from: Nick.C on
I iterated a few times until I just used 22223 (the same as for the 64 sample FFT) as increasing the 2's results in a higher bitrate(!). 22222 was also higher in bitrate - 22223 seems to be a sweet spot (some averaging, but not too much).
Thank you, Nick.
Do you mind making -spf temporarily available to the user again? (I'm only interested in playing around with the -spf setting for the 32 samples FFT, I'm just curious about the quality of 22222).
I'll post a beta v0.9.7b shortly.
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1144 –
Quote from: Nick.C on
Is the batch file on a path with spaces in it? I found this to be an elusive problem to solve.


I can send you a modified version that handles that, and unicode...  but yeah, some things are best left simple.  Let me know if I could help when I can.

lossyWAV Development

Reply #1145 –
Quote from: jesseg on
Quote from: Nick.C on
Is the batch file on a path with spaces in it? I found this to be an elusive problem to solve.
I can send you a modified version that handles that, and unicode...  but yeah, some things are best left simple.  Let me know if I could help when I can.
The problem seems to be with either foobar2000 or cmd.exe (I never did determine which). As soon as I removed spaces from the path to the batch file everything began to work.

Unicode handling in which sense?
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1146 –
Quote from: Nick.C on
Is the batch file on a path with spaces in it? I found this to be an elusive problem to solve. That is why my batch file is in a simple <drive>:\BIN\ directory, as are flac.exe and lossyWAV.exe - also, ensure that the batch file references the correct locations of the two relevant .exe files.

[edit2] Oh, and it's a lossy pre-processor which produces modified WAV files. It works with other codecs apart from FLAC, although I use FLAC by preference as it is compatible with TCPMP v0.81 on my iPAQ. [/edit2]

[edit] For example: flossy.bat
Code: [Select]
@echo off
c:\data_nic\bin\lossyWAV %1 %3 %4 %5 %6 %7 %8 %9 -low -nowarn -quiet
c:\data_nic\bin\flac.exe -5 -f -b 512 "%~N1.lossy.wav" -o"%~N2.flac" 
del "%~N1.lossy.wav"
with the batch file, lossyWAV.exe and flac.exe in the same directory, i.e. C:\DATA_NIC\BIN\ and called from foobar2000 with:

Encoder: cmd.exe

Extension: lossy.flac (NOT .lossy.flac!!!)

Parameters: /d /c c:\data_nic\bin\flossy.bat %s %d <insert your parameters here>

example parameters could be: -q 4 -impulse [/edit]


That's got it working!   
Thanks for the quick reply!
Batch files make me nostalgic for DOS 3.3  -- NOT!
Before I waste lots of time, assure me that there'd be no benefit taking a lossy.flac and converting it into MP3, AAC (HE, LC), OGG or some other lossy. (That's as an alternative to using a lower quality in the native encoder, thus relying on the inbuilt psychoacoustic tunings.)  I just started thinking about creating Nero LC-AAC files from your semi-lossies as an alternative to HE-AAC for my Sony-Ericsson musicphone...
Garf claims HE-AAC isn't battery-thirsty (though it is CPU hungry), but I have doubts.

lossyWAV Development

Reply #1147 –
Quote from: ckjnigel on
Before I waste lots of time, assure me that there'd be no benefit taking a lossy.flac and converting it into MP3, AAC (HE, LC), OGG or some other lossy. ...

When targeting at mp3, aac, ogg, etc. it's always best you encode from the original or a lossless codec.
If you use a high quality setting of lossyWAV (for instance for archiving instead of using a lossless archive) and convert from this to mp3, it is expected however that the quality loss due to this transcoding is insignificant.
lame3995o -Q1.7 --lowpass 17

lossyWAV Development

Reply #1148 –
Quote from: ckjnigel on
That's got it working!   
Thanks for the quick reply!
Batch files make me nostalgic for DOS 3.3  -- NOT!
Before I waste lots of time, assure me that there'd be no benefit taking a lossy.flac and converting it into MP3, AAC (HE, LC), OGG or some other lossy. (That's as an alternative to using a lower quality in the native encoder, thus relying on the inbuilt psychoacoustic tunings.)  I just started thinking about creating Nero LC-AAC files from your semi-lossies as an alternative to HE-AAC for my Sony-Ericsson musicphone...
Garf claims HE-AAC isn't battery-thirsty (though it is CPU hungry), but I have doubts.
Glad to be of service .

The added compression in certain lossless codecs is only due to the exploitation of the wasted-bits mechanism. Transcoding from a lossyWAV processed lossless files has not (to my knowledge) been well explored yet.

@halb27: I've been looking at making spreading functions for all of the FFT lengths more conservative. I'm trying: 22222-22222-22223-12233-12234-12234 and although the bitrate goes up a bit, it may be attractive.

[edit] 53 problem sample bitrates beta v0.9.8:
Code: [Select]
|----------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
| lossyWAV | -q 10 | -q 9  | -q 8  | -q 7  | -q 6  | -q 5  | -q 4  | -q 3  | -q 2  | -q 1  | -q 0  |
|----------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
| v0.9.8   |635kbps|609kbps|583kbps|556kbps|528kbps|500kbps|457kbps|419kbps|386kbps|358kbps|336kbps|
|----------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
| v0.9.8 i |644kbps|619kbps|594kbps|567kbps|539kbps|512kbps|469kbps|431kbps|399kbps|372kbps|351kbps|
|----------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
Looking at the bitrates however, it may be that this is too conservative. Advice / comment / opinion will be very well received.... [/edit]

[edit2]Trying 22222-22223-22223-12234-12234-12235, I get:
Code: [Select]
|----------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
| lossyWAV | -q 10 | -q 9  | -q 8  | -q 7  | -q 6  | -q 5  | -q 4  | -q 3  | -q 2  | -q 1  | -q 0  |
|----------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
| v0.9.8   |622kbps|596kbps|569kbps|541kbps|514kbps|487kbps|445kbps|408kbps|377kbps|351kbps|331kbps|
|----------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
| v0.9.8 i |635kbps|611kbps|585kbps|557kbps|530kbps|503kbps|462kbps|425kbps|394kbps|369kbps|349kbps|
|----------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|-------|
[/edit2]
lossyWAV -q X -a 4 -s h -A --feedback 2 --limit 15848 --scale 0.5 | FLAC -5 -e -p -b 512 -P=4096 -S- (having set foobar to output 24-bit PCM; scaling by 0.5 gives the ANS headroom to work)

lossyWAV Development

Reply #1149 –
Hallo Nick,

Bitrate increase of problem samples is welcome.
I always wonder in the first place what's the bitrate increase of regular music.
My personal opinion is that we should be very defensive towards the HF region with the short FFTs in the first place, and this addresses the 32 sample FFTs in case we consider these as something useful in the end.
Maybe it's a good strategy to leave the -spf setting for the standard analysis, but with a user supplied -analyses not just add one or more analyses but use a more defensive -spf setting for the added analyses.

Thank you for v0.9.7b. I'm curious about the bitrate with regular music and the quality of this version.
lame3995o -Q1.7 --lowpass 17