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: Lower algorithm quality in LAME produces better quality than anything above -q 3 (Read 10606 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #25
I have the same impression , may I ask you if you have this impression even 320 kb ?
[...]
Would like to perform/make some tests between 320k between 3.96.1 vs LAME 3.100.1 with -q0

I don't expect to be able to detect any difference at 320 kbit/s. That bitrate is just too high for my ears to reliably detect a difference.

Indeed, for "old Lame", I think 3.96.1 is a good version to look at.
Hi
well i will do some test between 96 and 192
thanks
Stick with v 3.97.  That is the last one to have ALL modes tuned inc CBR. There are listening tests that show
v3.97 has better in quality overall vs v3.90 incl cbr/abr for 128 k  .
hi
can I know where can I get 3.97 ? I can't find at rarewares
thanks

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #26
@shadowking

What about ABR mode in 256 - 288k range? Is it still good to use -q7 or standard -q3?
Also, should we stick with 3.97 on high bitrate (256+) or is version 3.100 equally good?

I like and use 3.100.1 rarewares as its very fast.  256 ..288 they should be perfect on normal music and acceptable on
extreme samples hopefully.  Sticking to -q3 should be fine and -q5 or more for speed if needed.  Can try --lowpass 17, -f or --resample 48 too.

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #27
I have the same impression , may I ask you if you have this impression even 320 kb ?
[...]
Would like to perform/make some tests between 320k between 3.96.1 vs LAME 3.100.1 with -q0

I don't expect to be able to detect any difference at 320 kbit/s. That bitrate is just too high for my ears to reliably detect a difference.

Indeed, for "old Lame", I think 3.96.1 is a good version to look at.
Hi
well i will do some test between 96 and 192
thanks
Stick with v 3.97.  That is the last one to have ALL modes tuned inc CBR. There are listening tests that show
v3.97 has better in quality overall vs v3.90 incl cbr/abr for 128 k  .
hi
can I know where can I get 3.97 ? I can't find at rarewares
thanks


http://web.archive.org/web/20070811105916/http://homepage.ntlworld.com/jfe1205/LAME/

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #28
If someone would like to compile a list of the various LAME versions that are required, I will do my best to make them available. :)

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #29
Hi everyone,

for testing purposes i compiled some of the LAME versions anew:
- Unmodified source code from SourceForge
- GCC 13.1 and NASM used
- Same CFLAGS across all compiles

Also attached is an encoding overview of these versions using -abr 192.
This is just to see the differences of e.g. the default “qval” values, encoding speeds, bitrate distribution, LR/MS ratio.

Martin

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #30
Looking at the attachment
--abr 192 =  --preset 192  from lame 3.94-3.100
before that --abr and -b used non preset system.
-q5 is default for 3.90 non preset and -q2 for preset.
-q2 is default for 3.93 both preset and non
-q2 mapped to -q3 from 3.94-3.100
3.97-3.100 are likely very similar.

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #31
The problem doesn't appear as obvious before lame 3.95 with either -b 128 or --preset 128.
I have come up with an experimental workaround using developer switches that need lame < 3.100 ;
In particular 'treble' and also 'alto' help a lot with highly tonal music.  For VBR the bitrate is increased up to
30 kbits for V2 . 

lame -b 128 --ns-bass -2.5 --ns-alto -2.5 --ns-treble -2.5
lame -V2  --ns-bass -2.5 --ns-alto -2.5 --ns-treble -2.5
lame -V1 -Y  --ns-bass -2.5 --ns-alto -2.5 --ns-treble -2.5

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #32
Hi everyone,

Martin
Hi Martin
may I know if you have compiled and can you compile them?
just because I have tried to compile many time without luck
thank you so much!

@shadowking

Quote
The problem doesn't appear as obvious before lame 3.95 with either -b 128 or --preset 128.
Hi
I haven't performed some serious tests but i have tried 3.93.1 and used my headphones and for my ears it does sound better the lower bitrate
may I know what do you mean for ?
Quote
lame --preset cd -f

it's in your signature


Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #33
 --preset cd is alias for --preset 192 or --abr 192


Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #35
@shadowking
Do you recommend using -f in general on high bitrate (ABR 256+) lame v3.100?

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #36
@shadowking
Do you recommend using -f in general on high bitrate (ABR 256+) lame v3.100?

I would say yes, From my testing it has cleaner sound overall on normal samples, good with EIG and other electronic, .
I find the -f distortions if any less objectionable than default settings most of the time .
There are a few specific distortions it adds on minority of artificial samples like this weird sample, florida_seq & serioutrouble @ 192k.  @ 256+ i might not be able to hear those.

I copied my collection to a USB stick for the car as well as sdcard for my phone. All encoded abr 192 -f .
I am quite pleased so far , nothing to arouse any suspicion.


Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #38
@john33
Hi
thank you so much!

@shadowking

Quote
I would say yes, From my testing it has cleaner sound overall on normal samples, good with EIG and other electronic, .
Hi
I have never used the parameter -f , the only information I can find is
Quote
Alias of -q 7
in the Aliases and removed settings list but there is informations
I will try even @96kb , have you performed some tests at 96kb?
thanks

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #39
@shadowking
Do you recommend using -f in general on high bitrate (ABR 256+) lame v3.100?

I would say yes, From my testing it has cleaner sound overall on normal samples, good with EIG and other electronic, .
I find the -f distortions if any less objectionable than default settings most of the time .
There are a few specific distortions it adds on minority of artificial samples like this weird sample, florida_seq & serioutrouble @ 192k.  @ 256+ i might not be able to hear those.

I copied my collection to a USB stick for the car as well as sdcard for my phone. All encoded abr 192 -f .
I am quite pleased so far , nothing to arouse any suspicion.

Great! It's even faster that way. :)

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #40
Hi Martin
may I know if you have compiled and can you compile them?
just because I have tried to compile many time without luck
thank you so much!

Hi Francesco,
i'm not quite sure if I've understood you correctly.
You want to know how I compiled the GCC binaries?

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #41
Hi Francesco,
You want to know how I compiled the GCC binaries?
Hi
i have tried to compile some lame binaries , i have never been able to do it
but now i don't need them ,just because there avaible in this topic
thanks

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #42
That was a bit of a struggle for me too - it's been a while  ;)
I don't think I could even remember every step to explain the whole process.

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #43
@john33
Hi
thank you so much!

@shadowking

Quote
I would say yes, From my testing it has cleaner sound overall on normal samples, good with EIG and other electronic, .
Hi
I have never used the parameter -f , the only information I can find is
Quote
Alias of -q 7
in the Aliases and removed settings list but there is informations
I will try even @96kb , have you performed some tests at 96kb?
thanks

Yes its better for me @ 96k

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #44
The LAME dev's made CBR and ABR use the same psychoacoustic model as VBR New in 3.99.0. Maybe this has impacted the quality. I've tried CBR 128, all default settings with 3.98.4 and it sounds better to my ears than in 3.99.5.

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #45
I think -f switch makes -V 5 or -V 6 sound way better when using --vbr-new in 3.100. The default -q 0 adds way too much artifacts
LAME: -f -V 0 -Y
 Xing: -V150 -X2 -U2 -HF-1 -TX0

 

Re: Lower algorithm quality in LAME produces better quality than anything above -q 3

Reply #46
I downloaded Fatboy sample of LAME 3.100  V0 and CBR 320 from here. https://hydrogenaud.io/index.php/topic,121579.0.html

320k CBR sounds really strange and worse than V0 or V2.  No need to do ABX as sample is really hard for all encoders. Here is a log of blind test between different encoders.

Code: [Select]
ABC/HR for Java, Version 0.53a, 20 October 2024
Testname:

Tester:

1L = LAME MP3 V0\13 fatboy_30sec.wav
2L = LAME MP3 192 kbps\13 fatboy_30sec.wav
3R = Apple AAC 192 kbps\13 fatboy_30sec.wav
4R = LAME MP3 320k\13 fatboy_30sec.wav

Ratings on a scale from 1.0 to 5.0

---------------------------------------
General Comments:
---------------------------------------
1L File: LAME MP3 V0\13 fatboy_30sec.wav
1L Rating: 4.4
1L Comment:
---------------------------------------
2L File: LAME MP3 192 kbps\13 fatboy_30sec.wav
2L Rating: 4.2
2L Comment: clipping
---------------------------------------
3R File: Apple AAC 192 kbps\13 fatboy_30sec.wav
3R Rating: 4.8
3R Comment:
---------------------------------------
4R File: LAME MP3 320k\13 fatboy_30sec.wav
4R Rating: 3.6
4R Comment:
---------------------------------------

ABX Results:

 Could it be related to this -q issue ?
Edit: Probably not, as files have been encoded with default -q 3 option. 
Edit 2 :Encspot shows  that 320 CBR has lower  "Big values" of bit reservoir.