LAME vs Helix speed discussion, split Options

[ddrawley]

I continue to be at a loss about the obsession over encoding speed. L.A.M.E. is a fantastic encoder. I encode my files ONCE. On a rare occasion, I reencode from my FLAC masters.
Then I listen, listen, listen.
Why would I potentially give up sound quality to achieve a twenty percent faster encoding speed? I encoded ONCE, and listened forever thereafter.
Call me a L.A.M.E. zealot if you will, I see no reason to use another encoder.
I give hearty thanks to all who have contributed to L.A.M.E. development.

[bug80]

I continue to be at a loss about the obsession over encoding speed. L.A.M.E. is a fantastic encoder. I encode my files ONCE. On a rare occasion, I reencode from my FLAC masters.
Then I listen, listen, listen.
Why would I potentially give up sound quality to achieve a twenty percent faster encoding speed? I encoded ONCE, and listened forever thereafter.
Call me a L.A.M.E. zealot if you will, I see no reason to use another encoder.
I give hearty thanks to all who have contributed to L.A.M.E. development.

That's exactly my opinion too.

But, I wouldn't be surprised if a lot of people think the opposite, they just want to encode their music to 128 kbs CBR as fast as possible for use on their portable. These kind of mp3 users won't visit HA.org quite often, however

[QuantumKnot]

Well, I often enjoy the speed of Xing for encoding audio for movies, since quality isn't always an issue with movies. After some very informal speed tests, it seems this new Helix encoder is faster than Xing 1.5. Encoding a 5 minute song took rougly 6.5 seconds using Xing and only 3.6 seconds using the new encoder. So it is much speedier

Has anyone done tests to compare the quality difference between Xing and the Helix mp3 encoder?

This post has been edited by QuantumKnot: Aug 4 2005, 13:54

[guruboolez]

... to achieve a twenty percent faster encoding speed?

20%?! It's rather 300...800% compared to LAME ABR/VBR/CBR (source: nyaochi).
I've tried myself with with portable computer. Reencoding from a lossless source gives me x35 speed whereas LAME is only ~x8.

I encoded ONCE, and listened forever thereafter.

Not me. I encode few albums for my portable, and I often permanently delete them. I don't have space enough to store my entire lossless collection + an entire MP3 library.

This post has been edited by guruboolez: Aug 4 2005, 14:32

[woody_woodward]

I continue to be at a loss about the obsession over encoding speed. L.A.M.E. is a fantastic encoder. I encode my files ONCE. On a rare occasion, I reencode from my FLAC masters.
Then I listen, listen, listen.
Why would I potentially give up sound quality to achieve a twenty percent faster encoding speed? I encoded ONCE, and listened forever thereafter.
Call me a L.A.M.E. zealot if you will, I see no reason to use another encoder.
I give hearty thanks to all who have contributed to L.A.M.E. development.

I don't disagree with you, but my situation is different. I download computer/internet/technology news to listen to while driving, or walking, or just eating lunch. I usually have to transcode from Real or WMA to mp3. This is speech, not music. I will normally listen to it one time. Also, I am not at all sure that LAME is superior at bit-rates of 24-32k. There is no doubt that it is slower. Xing seems a closer match to my needs. Having choices is a good thing.

[ddrawley]

Maybe someone needs to make a feature request of the L.A.M.E. team.
A new switch.
-encodereallyreallyfastidontcarehownastyitsounds

Would this be an -alt switch?



Good point on the choices aspect.

Edit: better switch name

This post has been edited by ddrawley: Aug 4 2005, 20:07

[Cygnus X1]

Maybe someone needs to make a feature request of the L.A.M.E. team.
A new switch.
-encodereallyreallyfastidontcarehownastyitsounds

It's called -q9.

[Zurman]

Maybe someone needs to make a feature request of the L.A.M.E. team.
A new switch.
-encodereallyreallyfastidontcarehownastyitsounds

Would this be an -alt switch?



Good point on the choices aspect.

Edit: better switch name

Some people keep only lossless files on their HD and encode their music each time they upload it on their flash mp3 player, which happens quite often due to the limited capacity (eg 256 Mb, around 5 albums with standard preset)

Those people may be interested in high speed and not in quality, considering the listening conditions (noisy environment, poor headphones, ...)

[HisInfernalMajes...]

Maybe someone needs to make a feature request of the L.A.M.E. team.
A new switch.
-encodereallyreallyfastidontcarehownastyitsounds

Would this be an -alt switch?



Good point on the choices aspect.

Edit: better switch name

Hehe, I can see it now "--alt-preset crap"

[LMS]

Speaking of crap, perhaps time to split off some of these OT posts to the recycle bin?

[rjamorim]

Speaking of crap, perhaps time to split off some of these OT posts to the recycle bin?

I agree. People are almost blaming others for daring using anything other than lame.

Bleh. Useless zealots.

[DigitalDictator]

I continue to be at a loss about the obsession over encoding speed. L.A.M.E. is a fantastic encoder. I encode my files ONCE. On a rare occasion, I reencode from my FLAC masters.

I'm thinking if Helix is almost on par with LAME (depending how you define 'almost') quality wise at this point, it might end up as a very good alternative with a few minor tweaks. It seems there are headroom for some tweaks and tunings, whereas LAME is in another situation where every single minor improvement comes from a handful of more grey hairs on both Gabriel and Robert.

Nobody's picking on the holy LAME, but some just find the high encoding speed favorable.

edit: redundant yada yada...

This post has been edited by DigitalDictator: Aug 5 2005, 10:42

[guruboolez]

I did a small and fast listening test between Helix CBR 128 and Gogo CBR 128, with Helix VBR (-V50) as third possible competitor (-V50 correspond to an average bitrate of 127 kbps on the 40 tested samples).

Comments are in french, but table and plots are understandable:

http://guruboolez.blogspot.com/2005/08/com...-mp3-ultra.html

[QuantumKnot]

Anyone tested Gogo vs Helix on the 'Waiting' sample?

[Gabriel]

I did a small and fast listening test between Helix CBR 128 and Gogo CBR 128, with Helix VBR (-V50) as third possible competitor (-V50 correspond to an average bitrate of 127 kbps on the 40 tested samples).

Interesting results.
At first glance they may seem to be in contradiction with the 128kbps mp3 test:
http://www.rjamorim.com/test/mp3-128/results.html

However, while looking more carefully, we clearly see that with the same samples, results are similar. Therefore, the different conclusion between your test and Roberto's one seems to be cause by the sample set used.

Now, the big question: Does that mean that the sample set in Roberto's test was too small? Same question for your test, if we only consider the "various" samples.

[guruboolez]

I haven't exactly tested the same encoders and/or settings

current test : gogo = -b 128 (v.3.13)
former test : gogo = -b 128 -a -q 0 (v3.12)
current test : helix = -V50
former test : Xing = VBR "quality normal"

additional note from Roberto:

# The Xing codec tested here is the last version available from XingTech. It was later bought by RealNetworks and underwent further development. The latest version - which wasn't tested here - is available with RealPlayer 10/RealOne

I suppose that helix encoder v.5 corresponds to the latest version which benefits from "further development".

In these conditions, it seems very "risky" to compare both results. Moreover, there are only four common samples between both tests:

CODE

sample     gogo-guru   gogo-coll   helix-guru-CBR   helix-guru-VBR   helix-coll

DaFunk        2,5         3,2         1,5               4,0              3,3
EnolaGay      1,0         3,6         1,3               2,0              3,3
experiencia   3,5         3,3         3,6               4,0              3,4
NewYorkCity   2,0         3,8         1,6               1,2              3,3

             2,25        3,47        2,00              2,80             3,32

Severity apart, results between tests are similar if we compare helix CBR to gogo CBR: gogo is slightly better. But we know that the collective tests used a VBR encoding of Xing, and ABR for Gogo.

This post has been edited by guruboolez: Aug 7 2005, 10:45

[kode54]

Now for more off-topic divergence and zealotry.

I can understand using a faster encoder to save time when encoding on the fly for a flash player, but since when did DVD ripping become a time critical process? Are we trying to go for a world record 0day screener release time or something? I realize the pressure for a privileged movie previewer to release a screener rip to the ftp top sites as fast as possible, but this is ridiculous.

[Gabriel]

but since when did DVD ripping become a time critical process

???
Who spoke of dvd ripping here?

[guruboolez]

QuantumKnot, here.

[Gabriel]

He mentionned encoding movies, not especially ripping dvds and releasing screeners.
I myself re-encode movies from TV to save space (I use my computer as a pvr, but it is not fast enough to do efficient real-time compression)

[guruboolez]

Exact. Sorry, in my mind digital movie corresponds to DVD, especially when people talk about ripping/reencoding. My mistake...
I haven't thought about this relevant usage of ultra-fast mp3 encoders. It's a good point for them.

This post has been edited by guruboolez: Aug 7 2005, 13:59

[robert]

If speed really matters, you can try:
-V4 --vbr-new -q7
-V2 --vbr-new -q7 -ms

[Defsac]

I store all my music with WavPack (-mhx6). When I use my portable, I batch convert the entire playlist (several albums) to MP3. In this situation speed differences are definitely noticeable, and it's a one time use encode (after I'm done I delete the lossy files and re-encode later), so it's not an 'encode once playback multiple times' scenario.

Personally I use LAME 3.97a11 with -vbr-new.