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: alt-preset new improvements? (Read 7492 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

alt-preset new improvements?

Hello.
I heard that --alt-preset will be improved in couple of weeks. What new is planned and how it compares to current settings?

alt-preset new improvements?

Reply #1 –
The improvements planned are mainly the following:

- Improve the ABR modes by implementing some of the same techniques used in the VBR modes
- Attempt to tune the VBR modes further to reduce artifacts present on a few samples
- Possibly create a new, lower bitrate VBR mode

I'm not sure exactly when this will be done (some of it already is done, though the work is incomplete), because I'm also trying to juggle a few other projects simultaneously, and these days at least, I'm a bit short on time.  I know some people are waiting on me to finish these up before they hold some public listening tests to compare ABR vs VBR though, so I'm trying to get things done as quickly as possible..

alt-preset new improvements?

Reply #2 –
* NeoRenegade drools.

Keep up the good work Dib!

alt-preset new improvements?

Reply #4 –
You settle down and do some of that good work mate!
It would be nice to see --a-p e a bit more tuned for transparency - not because I really trust VL, but I still think there should be some consideration.
Have you tried psy2 yet?
the 4096 FIR/FFT seems quite interesting ,and should yield in improved accuracy

cheers,
David

alt-preset new improvements?

Reply #5 –
Quote
Originally posted by MTRH
Have you tried psy2 yet?
the 4096 FIR/FFT seems quite interesting ,and should yield in improved accuracy


Does psy2 implementation already in LAME Project CVS? I just know that it has been mentioned in Shibata's homepage.

alt-preset new improvements?

Reply #6 –
Quote
Originally posted by Enig
Does psy2 implementation already in LAME Project CVS? I just know that it has been mentioned in Shibata's homepage.
Partly, you need an external program for pre-processing if you want to use nspsytune2.
Juha Laaksonheimo

alt-preset new improvements?

Reply #7 –
Quote
Originally posted by JohnV
Partly, you need an external program for pre-processing if you want to use nspsytune2.


I have two questions to ask:

1. where can I download this pre-pocessing binary?

2. what's the virture of nspsytune2 contrast with nspsytune?

alt-preset new improvements?

Reply #8 –
Quote
Originally posted by Dibrom
- Possibly create a new, lower bitrate VBR mode
How low is lower? I'd like to see how good quality you can squeeze out of a preset that behaves bitrate-wise like the old --r3mix (around 165kbps I think).

But I don't always get what I wish for.

In any case, low is good.  Even APS bloats on a lot of my metal (216kbps+).

alt-preset new improvements?

Reply #9 –
Quote
Originally posted by NeoRenegade
How low is lower? I'd like to see how good quality you can squeeze out of a preset that behaves bitrate-wise like the old --r3mix (around 165kbps I think).


I don't really know yet.  I'm not making any promises, but maybe something a bit more in that range will be possible.  The problem is that there are only so many sacrifices that can be made before quality starts to suffer too much.

Quote
In any case, low is good.  Even APS bloats on a lot of my metal (216kbps+).  


I don't think there will ever be a case where noisy music like metal will encode with a bitrate much lower than 192kbps with MP3 at least while still trying to maintain a high level of transparency.  This is mainly just due to the inefficiencies present within the format which have already been discussed many times before.

If bitrates at this level are a bit of a problem, I seriously suggest looking at a different format, because no matter how many tweaks I make in this regard, some things just aren't possible  At least, unless you can change the rules so to speak.. but with MP3 we can't really do that.

alt-preset new improvements?

Reply #10 –
could there be some kind of 2pass vbr encoding scheme ?

Woulde be really god for video encoding...backups you know :-)

something like --2pass xxx where XXX is the average values

so far to come NEAR a 2pass vbr encoding i encode with abr --alt-preset xxx

then i se how much is is of, and add to the value and encode again.
Then i hit the exact bitrat +/-1 kbits

woulde be nice to be able to do this automatic in one command line
Sven Bent - Denmark

alt-preset new improvements?

Reply #11 –
Quote
Originally posted by Enig
I have two questions to ask:

1. where can I download this pre-pocessing binary?
I think you have to contact Naoki. This is probably because nspsytune2 is in heavy development stage, and it should be only used for testing..
Quote
2. what's the virture of nspsytune2 contrast with nspsytune?

- Better tonality estimation.
- Psychoacoustics will use 4096 point hybrid FIR/FFT filter to improve frequency resolution of psychoacoustics. (Without nspsytune2 Lame uses 1024 point FFT)
- Possibly soon: better temporal masking to help reduce pre-echo.
- Possibly soon: other short block bit allocation improvements (to help reduce pre-echo).
Juha Laaksonheimo

alt-preset new improvements?

Reply #12 –
I'm lost and spamming this forum...What's a good command line for encoding mostly metal? the speed is unimportant and quality is the main issue. Would a CBR 320 be the only choice (large file sizes and all)?
I\'m new at this...

alt-preset new improvements?

Reply #13 –
Quote
Originally posted by astro
I'm lost and spamming this forum...What's a good command line for encoding mostly metal? the speed is unimportant and quality is the main issue. Would a CBR 320 be the only choice (large file sizes and all)?

Try:

--alt-preset extreme

Cheers.

alt-preset new improvements?

Reply #14 –
Quote
Originally posted by NeoRenegade
How low is lower? I'd like to see how good quality you can squeeze out of a preset that behaves bitrate-wise like the old --r3mix (around 165kbps I think).

 



Yes! I am looking for a VBR setting around 160Kbps for use in portable (flash mem) player.  160 ABR would probably do, but wouldn't a VBR setting delivering the same rate (on average) do a better job?

The current presets give me the impression that VBR cant be trusted when the lowest one (standard) encodes my Rush albums at an Average of 220 Kbps.

I realize that there would be some loss of transparency with a VBR setting that delivered 160 Kbps on the same track, but surely it should be better than 160 ABR/CBR ( which look much the same to me).

Peter

alt-preset new improvements?

Reply #15 –
>> - Possibly create a new, lower bitrate VBR mode
That would be nice, and useful.  Somthing like radio and voice would be very nice.

(My dad is interesting in encoding some speaches into MP3).

alt-preset new improvements?

Reply #16 –
Quote
Originally posted by guidryp

Yes! I am looking for a VBR setting around 160Kbps for use in portable (flash mem) player.  160 ABR would probably do, but wouldn't a VBR setting delivering the same rate (on average) do a better job?

No. If you have certain bitrate in mind, it is preferable to use ABR because (a) you don't have to guess what -V
  • setting to use to generate a file of a certain bitrate, (b) ABR is rather consistent and you don't have to worry about psychomodel issues with VBR (like using too low of a bitrate on some frames), © ABR encodes a bit faster than VBR.

alt-preset new improvements?

Reply #17 –
Quote
Originally posted by mithrandir

No. If you have certain bitrate in mind, it is preferable to use ABR because (a) you don't have to guess what -V
  • setting to use to generate a file of a certain bitrate, (b) ABR is rather consistent and you don't have to worry about psychomodel issues with VBR (like using too low of a bitrate on some frames), © ABR encodes a bit faster than VBR.



So let me get this straight. If I made the effort or by trial and error and encoded with a -V setting that resulted in a similar bit rate for a specific song to ABR 160, would the VBR not produce a better sounding file?

Or is VBR only useful for 200Kbps + encoding?

Peter

alt-preset new improvements?

Reply #19 –
Quote
Originally posted by tangent
Reading this might give you an idea:
http://www.hydrogenaudio.org/forums/showth...s=&threadid=984
It is well known that at low and medium bitrates up to 160kbps, there is no VBR setting which is better than the ABR setting at equivilent, especially with the ABR --alt-presets now available since Lame v3.90.


Interesting read, but I still question the statement that up to 160 there is not VBR setting which can match ABR (which appears little different than CBR). Is this test still pending?

Again I will state that this implies to me that there are issues with the VBR models if they can't compete with very simplistic almost CBR algorithm.

I guess I missed the disclaimer on the statement that VBR was better than ABR was better than CBR. Disclaimer would be what?Only at high bit rates above 192kbps???

Is there another encoder out there which offers better mid-low VBR performance?

Peter

alt-preset new improvements?

Reply #20 –
Quote
Originally posted by guidryp


Interesting read, but I still question the statement that up to 160 there is not VBR setting which can match ABR (which appears little different than CBR). Is this test still pending?

ABR is *not* "little different than CBR."  It is a variable bitrate format, and it does exactly what you're talking about (keep bitrates around what you want them to be, in this case ~160k). 

So why don't you go with what's clearly appropriate in your case... use ABR.  The reason regular VBR is 'superior' is exactly because bitrates can go up as high as needed (very high in certain cases) -- and that's what you're arguing VBR shouldn't be doing... think about it.

alt-preset new improvements?

Reply #21 –
Quote
Originally posted by guidryp
Again I will state that this implies to me that there are issues with the VBR models if they can't compete with very simplistic almost CBR algorithm.

That's like complaining that a 4 cylinder engine cannot get 13 seconds in the 1/4 mile, like a V8 would, without considering that the 4 cylinder gets far better gas mileage and costs less. VBR is not the silver bullet and is suited to certain tasks such as high-bitrate/quality-centric encoding. But don't want high-bitrate. You said you want 160kbps. Therefore, don't worry about VBR for the moment because it's not the right tool for the task. It's like bringing a knife to a gunfight.

alt-preset new improvements?

Reply #22 –
First, I must admit that I am among those who are stuck thinking that VBR should always be better than ABR for the same reason that VBR is better than CBR. Now bear with me, I accept that this is not currently true with LAME, but I still think it should be, and I have been pushing for a lower bitrate VBR preset for a long time now.  I've kept my posts here to a very minimum, restricting myself to just browsing and learning for the last few months.

I have an idea now that takes a different approach to satiating listeners like myself afflicted with the above way of thinking/needs:
Perhaps instead of looking to tweak VBR to get lower bitrates, there needs to be some tweaking of ABR to let it flex more.  Granted, I haven't done any coding on LAME, but here's the way I see it working(for the end user):  You add a switch to your ABR mode if you want to get more "bitrate flex", perhaps there could even be a parameter to this switch that would give the amount of "flexing" with 0 being the default and acting exactly like ABR modes.  I could see the highest values perhaps being overly aggressive and "degressive" with bitrates, but maybe a setting of 2 or 3 would satisfy those of us who have a bitrate in mind, but aren't looking for that bitrate specifically.  Then, using something like --alt-preset 160 -flex 3 such tracks as standup comedy would come out ~105kbps, while metal tracks might average  ~195, and your typical pop-rock track ~150-170.

comments?

alt-preset new improvements?

Reply #23 –
I agree, with this condition -- that the "old" ABR modes be kept as-is.  ABR means "Average Bitrate," and once it starts flexing very much it isn't ABR anymore, by definition.  Or at least, the specified average bitrate loses a lot of meaning.

So my suggestion would be to relegate the "inflexible" ABR method (which is what ABR *is*, a specified average) to a new switch, say:

--alt-preset abr <xxx>

Then tweak flexibility, using something like what you described above.

alt-preset new improvements?

Reply #24 –
This kind of more flexible ABR would need more precise Perceptual Entropy algorythm. Infact, because of SFB21 issue of VBR coding, a more sensitive PE algorithm could be useful for about 160kbps targetting in order to allow the bitrate to flux more but still give more precise bitrate than pure VBR. This kind of "ABR"-mode allows high frequency coding without bloating, and would flex more especially towards higher bitrates when needed.

Not sure how easy this would be to implement.. Have to ask Robert, Naoki or Takehiro.
Juha Laaksonheimo