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: LAME -V 0 --vbr-new killer sample (Read 13815 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

LAME -V 0 --vbr-new killer sample

I found a problem sample on track last night and i find it transparent on 192 kpbs CBR by using --preset cbr 192 but its not transparent on -V 2, -V 1 and -V 0 --vbr-new on both LAME 3.97 and LAME 3.98b5.

LAME 3.98b5 -V 0

Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.4
2007/10/13 16:11:41

File A: F:\Listen Tests\Die In A Crash (Sample).wav
File B: F:\Listen Tests\Die In A Crash (Sample) LAME3.98b5 V0.mp3

16:11:41 : Test started.
16:12:09 : 01/01  50.0%
16:12:22 : 02/02  25.0%
16:12:30 : 03/03  12.5%
16:12:42 : 04/04  6.3%
16:13:20 : 05/05  3.1%
16:13:50 : 06/06  1.6%
16:14:08 : 07/07  0.8%
16:14:28 : 08/08  0.4%
16:14:42 : 09/09  0.2%
16:14:52 : 10/10  0.1%
16:15:16 : 11/11  0.0%
16:16:13 : 12/12  0.0%
16:16:16 : Test finished.

 ---------- 
Total: 12/12 (0.0%)


A nasty precho at around 0:15 - 0:18

All the vbr encoded versions of that track produce a huge bitrate bloat at around 245 on V2 and 263 on V0 thanks to that awfull sfb21 bug while the 192 cbr encode is transparent.

I have uploaded the lossless sample and the LAME 3.98b5 encoded mp3.

Lossless Sample in FLAC:

http://www.sendspace.com/file/19f5s6

Lossy sample in LAME 3.98b5 -V 0:

http://www.sendspace.com/file/iapt6f
"I never thought I'd see this much candy in one mission!"

LAME -V 0 --vbr-new killer sample

Reply #1 –
can you try also --vbr-old
 

LAME -V 0 --vbr-new killer sample

Reply #2 –
If it is indeed a problem sample for 3.98b5...looks like there will be a 3.98b6...
Zune 80, Tak -p4 audio library, Lossless=Choice

LAME -V 0 --vbr-new killer sample

Reply #3 –
Quote from: shadowking on
can you try also --vbr-old

I tried -V 0 --vbr-old on LAME 3.98b5 and i find it transparent unlike --vbr-new.
"I never thought I'd see this much candy in one mission!"

LAME -V 0 --vbr-new killer sample

Reply #4 –
you should be hired by Lame devs to spot artifacts!!!

LAME -V 0 --vbr-new killer sample

Reply #5 –
OT: Is there a Wiki Article with all samples? With links to download sources and relevant discussions? If not, would be nice to have a sample resource page...

LAME -V 0 --vbr-new killer sample

Reply #6 –
Quote from: Bourne on
you should be hired by Lame devs to spot artifacts!!!


Yeah that would be a neat job

Anyway I did a quick test with LAME 3.90.3 --alt-preset and its seems transparent to me, but its very easy to abx at LAME 3.97 and 3.98b5 at -V 0 --vbr-new and my hearing is not that great at all.
"I never thought I'd see this much candy in one mission!"

LAME -V 0 --vbr-new killer sample

Reply #7 –
Wow, my hearing isn't as good as I thought.

I can't hear the flaw.

I'm not quite sure what pre echo is.

LAME -V 0 --vbr-new killer sample

Reply #8 –
Quote from: ech3 on
I'm not quite sure what pre echo is.

Pre-echo is the opposite of a very short normal echo: it happens before the trigger instead of afterwards. For example, a sharp percussion hit, may sound more soft and "smeared". Visually, you could imagine it like an added quick "fade-in" of a sound.

Spotting artifacts has less to do with hearing/equipment and more with training. It is a common myth, that recognizing mp3-artifacts is mainly an issue of hearing and equipment. However, some people are more sensitive to certain artifacts than to others. I for example dont care that much about pre-echo and instead more about other artifacts - mainly those which happen at high frequency ranges (10-14khz).
I am arrogant and I can afford it because I deliver.

LAME -V 0 --vbr-new killer sample

Reply #9 –
Woohoo, I could make it on CMI8738 chip and with TV background noise.
Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.5 beta 1
2007/10/14 00:10:18

File A: D:\Die_In_A_Crash.flac
File B: D:\Die_In_A_Crash_V0.mp3

00:10:18 : Test started.
00:12:28 : 01/01  50.0%
00:13:23 : 02/02  25.0%
00:13:54 : 03/03  12.5%
00:16:25 : 04/04  6.3%
00:17:06 : 04/05  18.8%
00:17:34 : 05/06  10.9%
00:20:10 : 06/07  6.3%
00:22:19 : 07/08  3.5%
00:22:21 : Test finished.

 ---------- 
Total: 7/8 (3.5%)
stimulating the audio nerve directly

LAME -V 0 --vbr-new killer sample

Reply #10 –
I have never favored VBR at very high bitrate, but until today I did not think that CBR 192 is better than -V0 towards pre-echo on a sample, especially as 3.98b5 so far seems to have improved on pre-echo behavior.
Good to see that 3.98b5 --vbr-old is fine, as well as that  cbr is fine which I guess is valid for the more interesting abr mode as well.
lame3995o -Q1.7 --lowpass 17

LAME -V 0 --vbr-new killer sample

Reply #11 –
Quote from: halb27 on
I have never favored VBR at very high bitrate, but until today I did not think that CBR 192 is better than -V0 towards pre-echo on a sample, especially as 3.98b5 so far seems to have improved on pre-echo behavior.
Good to see that 3.98b5 --vbr-old is fine, as well as that  cbr is fine which I guess is valid for the more interesting abr mode as well.


ATM i recken -V 1 would be a good setting then -V0 since -V0 is only abit or not better then the -V1 and a 1mb bigger, most of time sticks upto 260 on most tracks i tried while V2 tracks go to 240 like 260 was max but I did get a track 300kbps    on V0 which is a sbf21 bug though. And i could have picked up artifacts on a 192 encode if I was trained to spot artifacts and I still belive that VBR is still better then 192 CBR and this sample is a bug on --vbr-new and the other tracks i abxed this week.

Quote from: yandexx on
Woohoo, I could make it on CMI8738 chip and with TV background noise.
Code: [Select]
foo_abx 1.3.1 report
foobar2000 v0.9.4.5 beta 1
2007/10/14 00:10:18

File A: D:\Die_In_A_Crash.flac
File B: D:\Die_In_A_Crash_V0.mp3

00:10:18 : Test started.
00:12:28 : 01/01  50.0%
00:13:23 : 02/02  25.0%
00:13:54 : 03/03  12.5%
00:16:25 : 04/04  6.3%
00:17:06 : 04/05  18.8%
00:17:34 : 05/06  10.9%
00:20:10 : 06/07  6.3%
00:22:19 : 07/08  3.5%
00:22:21 : Test finished.

 ---------- 
Total: 7/8 (3.5%)


Kinda glad to see someone else to abx that with background noise, like when my little brother was pestering me when i was abxing that track
"I never thought I'd see this much candy in one mission!"

LAME -V 0 --vbr-new killer sample

Reply #12 –
My impression is that CBR / ABR is inferior to VBR with preecho based on past tests.

LAME -V 0 --vbr-new killer sample

Reply #13 –
Hi i am a new here, so i keep me offside. But if someone finds a new sample that lame doesnt convert good without transparancy, do you guys want to know about it. Or do you have enough samples already?

LAME -V 0 --vbr-new killer sample

Reply #14 –
There is no such thing as an ever increasing collection of samples where Lame behaves badly.
The contrary is true: the Lame devs do a hard job improving on the issues found, and they do succeed - though perfection is not possible.
Any new sample that shows weaknesses of a specific encoder is welcome here, so there is a chance the encoder can be improved.
lame3995o -Q1.7 --lowpass 17

LAME -V 0 --vbr-new killer sample

Reply #15 –
Quote from: halb27 on
There is no such thing as an ever increasing collection of samples where Lame behaves badly.
The contrary is true: the Lame devs do a hard job improving on the issues found, and they do succeed - though perfection is not possible.
Any new sample that shows weaknesses of a specific encoder is welcome here, so there is a chance the encoder can be improved.

A few years back i did some testing how high my vbr should be before i could not heard the difference. I used the first  10 seconds from dayton - the sound of music. It has a very sparkling and high beginning. I listened with a headphone and i  had to up the bits (VBR HQ) each run, it sounded better each time, but i still could hear the difference between 320 cbr and the original cd. So thats why i remembered it. That was an older encoder dont know how it sounds now.

I tried to find the cd, but could not find it to test. If i find it, i come back on it. It must be somewhere.
If not then i hope someone else has it lying and some spare time  :-)