Hello,
I wrote mi question already in the title: Will Garf (or somebody else) present a highbitrate-tuned version of oggvorbis 1.01? I think it would be very practicle, so I don't have to decide between the two versions everytime.

Big_Berny

The tunings done in 1.0.1 mostly apply to low bitrates (<128kbps) while Garf's tunings apply to high bitrates (>160kbps); additionally Garf himself stated that he does not recommend using his GT versions for anything but the tuned qualities, since anything below might have been affected by his "Floggy" (~10kbps) tunings.

You'll have to wait for Monty to integrate further high bitrate tunings (based on GT3b1?) in Vorbis 1.1. Until then you are best off using GT3b1 (only) for quality settings >= 5.

It should be possible to take the 1.0.1 code and insert the tunings for high bitrate only, I will consider it.

Any reason not to? Just wondering...

Garf said (on IRC - as far as I understood him), that he might do a 1.0.1 based GT3 without the floggy hacks, which were included in GT3b1, so this version will be safe to use at all bitrates/quality settings.

This is great news. That means, we dont need to carry two encoders around all the time. Swapping those CDex DLLs around is a pain, you know

yes!

libvorbis-1.0.1+gt3b1 would be fantastic!

I dunno why that wasn't thought of before in the first place! PLEASE, roll that baby out as soon as possible!

Much better to have one super encoder out there right now listed on Rarewares, Foobar encoder plugin, etc.

So Garf's attempts at floppy-fitable Oggs are present in GT3b1? That would interest me. ^^

Ogg Vorbis GT3 beta 1





However, this would not be included in the next GT3 release, since it breaks some of the code for qualitities < 5, which is the reason GT3b1 is not recommended for these settings.

dev0

Can we have better floggy supported GT3 thanks. maybe Garf is not using it but I would and i have 7600 file to encode.

I, for one, would love to see Floggy tweaked & tuned. The sound quality is nowhere near grand, but such small filesizes would have very practical applications.

Not, say, for streaming. But for full-song demos that no-one could realistically argue are near CD quality. And immense bandwith savings, and a good initiative still left for most people to buy the full-quality song.

So...

Will Garf make this happen? If anything it would save a tier on the ogg section on rarewares where they each have a separate version for oggdropXPd.

It will save Rarewares some server space AND offer us one convienient solution for all our vorbis needs.

OK guys. This is for testing ATM. I think I have got it right, but I'd like a few people to confirm this before I publish it at Rarewares.

So, you're wondering, what the hell am I talking about??? oggdropXPd v1.7.5 compiled using libvorbis 1.0.1 BUT with GT3b1 integrated.

So would a few kind people d/l this (http://homepage.ntlworld.com/jfe1205/OggVo...GT3b1-1.0.1.zip) and let me know whether you think I have this right. My own testing indicates that the lower 'q' settings give the same results as 1.0.1 and the higher settings match GT3b1, but I'd like confirmation please.

TIA.

It would be really cool if the average bitrates of the GT3 version would be matched to the bitrates of the regular version at a given quality setting. I have no idea what amount of work is needed to achieve this.

I agree, but I'm afraid that's way beyond my capabilities!!

Someone really should test for john, too bad I don't know jack about vorbis.

Downloaded, and testing at the moment (but got to go out for an hour now!)

Here are some initial results, based on a couple of Foo Fighters tracks. All were encoded from flac files, encoding two at a time, copying tags from the flac file:


"Disenchanted Lullaby"
1.0.1 @ -q2.22 - 3193kb
johns @ -q2.22 - 3191kb

garf @ -q5 - 6393kb
john @ -q5 - 6408kb


"Low":
1.0.1 @ -q2.22 - 2952kb
johns @ -q2.22 - 2952kb

garf @ -q5 - 5819kb
john @ -q5 - 5839kb

So there are some size differences between the two, which suggests a possible problem. I didn't try ABXing the various samples, but all sounded OK - no obvious corruptions etc.

Cheers, Paul

Thanks, Paul. I think you'll find that the extra size of the higher bitrate, ie., the GT3b1 lib, encodes is actually down to the fact, IIRC, that the latest libogg creates slightly larger files that the version used in the 1.0-Gt3b1.

John,

I guessed it might be something like that! I have three albums in my queue to oggify, so I'll encode and listen to those this afternoon and post with my impressions (not set up for ABXing, but might reveal some issues).

You might want to repost your request as a new topic - I'm sure there are many more people interested in this than are reading this thread!

Cheers, Paul

@Paul,

You're probably right!! I'll wait for your feedback first though, I think!!

Listened to a Foo Fighters album at -q5 and a Carter USM one at -q4.26, and both sounded fine - no corruption, skipping etc, and from a limited comparison no obvious deviations from the source. Without a proper ABX that doesn't mean that you got the settings right, but it suggests that you didn't mess up too badly either

Thanks for the build!

Cheers, Paul

Thanks, Paul. I'll post a set of compiles at Rarewares shortly, and we'll see what happens.

Just what I've been waiting for. Thanks much John (and Garf)!

True. I merged the GT3b1 code into the libvorbis 1.0.1 tree some time ago and noticed a slight difference in file size. Haven't traced the source, though.

( http://matchasm.student.utwente.nl/libvorb...20031130.tar.gz )

Damien Rice - Delicate - Damien Rice Live From the Union Chapel

q 5.50 (original, listed as 176kbps) - 5.43MB (142kbps according to foobar)
q 5.50 (GT3b1, listed as 196kbps) - 5.94MB (155kbps according to foobar)

Both sound fine, but this is far from a 'killer sample' laden song, and the ALC650 on my work box is not sufficient for ABXing.

Of course this is just testing the non-GT3b1 version of oggdrop versus the hybrid version...