I've uploaded a new version of my tuned Ogg encoder.
New
-----
- Fixes 'nominal bitrate=999' issue in 160kbps mode
- tuned 350 kbps mode
- 'sortof' tuned 128kbps mode
http://sjeng.org/ftp/vorbis/oggdropgt2.exe (http://sjeng.org/ftp/vorbis/oggdropgt2.exe)
I'll put out a Linux tarball (and perhaps an oggenc) when I have more time.
--
GCP
...tune OGG for 40-48 Kbps (32KHz, mono).
Thanks
We're getting close to RC3?
What I'd like to hear is if ogg compares to mpc at around 170, 180 VBR ...
I know it's better than anything at lower bitrates, but how does it compete against mpc above 160kbps VBR?
Originally posted by PatchWorKs
...tune OGG for 40-48 Kbps (32KHz, mono).
Thanks
I think RC2 actually already includes such a mode by default, but I don't know if it's enabled in the actual encoder.
When I tested it, 48kbps 32kHz
stereo sounded pretty good, keeping in mind it was 48kbps.
48kbps 32kHz mono should be no problem. Doesn't the standard RC2 encoder have it?
--
GCP
Originally posted by lucpes
We're getting close to RC3?
I think so, although my tuned modes are still based on RC2 + some fixes. RC3 incorporates some of the improvements in the tuned modes into the default encoder.
What I'd like to hear is if ogg compares to mpc at around 170, 180 VBR ...
I know it's better than anything at lower bitrates, but how does it compete against mpc above 160kbps VBR?
The tuned 160kbps VBR mode is an attempt to provide an alternative to MPC. It doesn't get the same quality level as MPC does at that bitrate, but it's still pretty good. Beating MPC at it's forte will be hard.
--
GCP
Thanks for the quick answers... anyway I must say that the builds that have 128kbps limit beat r3mix mp3s...
Good thing that they're limited, 'cause I would start using them instead of mpc... this way I'll have to wait until rc3.
and again the same question
could you compile an oggenc.exe with the new tweaks?
Originally posted by Garf
I think RC2 actually already includes such a mode by default, but I don't know if it's enabled in the actual encoder.
True, but the encoder (OggEnc v0.8) says: "... the encoder is only tuned for rates of 44.1 and 48 kHz and while other rates will be accepted quality will be significantly degraded."
When I tested it, 48kbps 32kHz stereo sounded pretty good, keeping in mind it was 48kbps. 48kbps 32kHz mono should be no problem.
I have encoded the same wav file (32KHz, mono) through dbpoweramp and was successfull.
Doesn't the standard RC2 encoder have it?
dbpoweramp uses rc2, so the answer is yes.
I change the question in other way:
]i can encode at any KHz/Chans/Bits for sure, but i really want know which of them are more tuned... thanks
Originally posted by Benjamin Lebsanft
and again the same question
could you compile an oggenc.exe with the new tweaks?
Done. It's up on http://sjeng.org/ftp/vorbis (http://sjeng.org/ftp/vorbis)
--
GCP
Originally posted by PatchWorKs
True, but the encoder (OggEnc v0.8) says: "... the encoder is only tuned for rates of 44.1 and 48 kHz and while other rates will be accepted quality will be significantly degraded."
[...]
i can encode at any KHz/Chans/Bits for sure, but i really want know which of them are more tuned... thanks
The 'only tuned for 44.1/48khz' warning is outdated I think.
Edit: Segher corrected me on this. The encoder is really tuned for 44.1/48khz audio right now. Although lower sampling rates may usually work fine, they will be sensitive to preecho issues and such.
--
GCP
Thx Garf
It seems to me, that in 350 kbps mode there is the 19 kHz lowpass. Must it be at that high bitrate?
How does Vorbis compare to MPC at 256 - 350 kbps modes?
Originally posted by elfin
It seems to me, that in 350 kbps mode there is the 19 kHz lowpass. Must it be at that high bitrate?
There is no lowpass. It's just that the ATH+masking decides that most of those sounds aren't audible. That's what psychoacoustic compression is all about...
Edit: I checked, and my mode does really encode up to 20kHz if needed, and very sometimes even above that...
--
GCP
Originally posted by Garf
There is no lowpass. It's just that the ATH+masking decides that most of those sounds aren't audible. That's what psychoacoustic compression is all about...
--
GCP
Isn't the ATH+masking just a little too aggressive at such a high bitrate?
How about adding a switch to manipulate ATH & masking threshold?
Originally posted by elfin
Isn't the ATH+masking just a little too aggressive at such a high bitrate?
Unless you come up with a clip that you can positively ABX because of a lack of HF, the answer is no.
--
GCP
Originally posted by Garf
Unless you come up with a clip that you can positively ABX because of a lack of HF, the answer is no.
--
GCP
Though I have a pair of elven[/b] ears, a HQ amp and my loudspeakers weighing 90 kg, I can't ABX at that bitrate.
How does Vorbis compare to MPC at 256 - 350 kbps modes?
Modes b256 and b350 in RC2 still have some issues, pre-echo mostly. But Garf's tuned b350 really rocks - I tested it with several hard-to-encode test cases, like castanets and fatboy, and there's no pre-echo at all. Actually, I can't find a sample that has artifacts. The bitrate is roughly 300-350kbps, depending on the material you're encoding. I get around ~320kbps for most rock/metal stuff. I think this mode is very high quality and can be used for archiving. Since MPC is also transparent at these bitrates, I think there's no difference between Vorbis (i.e. Garf's tuned b350 mode) and MPC in 300+ kbps range. Definitely better then MP3 with same bitrate.
Great job, Garf. I'm very happy with your tuned -b350 mode
Heck, I even refuse to ABX the 160 kb/s mode. Absolutely too transparent.
Short while ago I whined about rc2's 160kb/s mode sounding horrible at this certain Test Clip, in fact worse than 128kb/s mp3.
Now what is really disturbing here is that the official rc2 oggenc @ 160 tries to maim my Official Test Clip in to a 179kb/s package.
GTuned2 manages to bring peace of mind and smooth harmony at 164kb/s. Serious progress here!
OGG rules
Thanks Garf!
Originally posted by elfin
[B9Though I have a pair of elven ears, a HQ amp and my loudspeakers weighing 90 kg, I can't ABX at that bitrate.
Outing myself : Mine weigh 65 kg each, tube/MOSFET amp, but i simply refuse to torture them with compressed music, even 24/96 is not good enough for my precious music IMHO ...
Compression, whatsoever kind, is good for my soundtracks in my DivX movies, but thats it ....
Originally posted by ChristianHJW
Mine weigh 65 kg each, tube/MOSFET amp, but i simply refuse to torture them with compressed music, even 24/96 is not good enough for my precious music IMHO ...
Compression, whatsoever kind, is good for my soundtracks in my DivX movies, but thats it ....
So what do you listen to? Live music only?
My loudspeakers are my own design, made after studying the loudspeaker design theory. None of the retail loudspeakers were good enough. Not even the expensive ones, which I could not afford anyway.
Now aren't we off topic? Sorry
to get back to the topic
as it is always me, Garf could you compile an oggenc.exe with the rc 2.9 release ??
Originally posted by Benjamin Lebsanft
to get back to the topic
as it is always me, Garf could you compile an oggenc.exe with the rc 2.9 release ??
That 'RC2.9' is a special version I did for the BBC.
It takes in 32kHz 16bit stereo and puts out 48kbps files, suitable for streaming over a modem.
You really want _that_ ?
--
GCP
ok, not really I'll stick with the gt2 version
hi garf.
in rc2, if not a wavelet or wavelet package transform is used?
if used, how deal with blocks connection between mdct and wavelet ?
regards.
Originally posted by Garf
That 'RC2.9' is a special version I did for the BBC.
It takes in 32kHz 16bit stereo and puts out 48kbps files, suitable for streaming over a modem.
You really want _that_ ?
--
GCP
YYYYYEEEEEESSSSSSSSSSS !!! I REALLY WANT IT
I need 32 kHz, 16 bit, mono please (bitrate: min 32 - max 48 - average 40 kbps)...
Tnx
Originally posted by jianxin yan
hi garf.
in rc2, if not a wavelet or wavelet package transform is used?
Nope. Wavelets aren't and won't be in Vorbis for a while.
Probably after 1.0 final release testing with them will begin.
--
GCP
Originally posted by PatchWorKs
YYYYYEEEEEESSSSSSSSSSS !!! I REALLY WANT IT
I need 32 kHz, 16 bit, mono please (bitrate: min 32 - max 48 - average 40 kbps)...
Tnx
Sorry, I've only got a stereo mode for now
I'll compile an oggenc when I have time.
(btw. It's 45kbps now, I shaved off some. Going to 40kbps average is probably not going to be listenable with 32kHz stereo.)
--
GCP
Test this file (open with WinAmp CTRL+L) and tell me what you think about quality:
http://www.patchworks.it/stream.ogg (http://www.patchworks.it/stream.ogg)
[encoded by Peter Pawlowski's Ogg Vorbis encoder plug-in v1.31 (libvorbis v1.0 RC2)]
Originally posted by PatchWorKs
Test this file (open with WinAmp CTRL+L) and tell me what you think about quality:
http://www.patchworks.it/stream.ogg (http://www.patchworks.it/stream.ogg)
[encoded by Peter Pawlowski's Ogg Vorbis encoder plug-in v1.31 (libvorbis v1.0 RC2)]
Hard to tell. First impression is quite good, but the song is not very stressing. As soon as some more instruments kick in you see the bitrate go up too.
Feed it some metal
--
GCP
- ORIGINAL FILE CHARACTERISTICS -
Length (m:ss): 1:30
Size: 5.799.592
Frequency: 32000 Hz
Channels: 1
Artist: BirdFlesh (http://home.swipnet.se/~w-92019/ (http://home.swipnet.se/~w-92019/))
----------------------------------------------------------------------
ENCODER = Ogg Vorbis encoder plug-in
File: test1
Location: http://www.patchworks.it/test1.ogg (http://www.patchworks.it/test1.ogg)
Average Bitrate: 46 Kbps
Size: 532.091
Compression Ratio: 10 to 1 (9 %)
ENCODER = dBpowerAMP Music Converter
File: test2
Location: http://www.patchworks.it/test2.ogg (http://www.patchworks.it/test2.ogg)
Average Bitrate: 43 Kbps
Size: 499.166
Compression Ratio: 11 to 1 (8 %)
Both encoders uses Xiphophorus libVorbis I 20010813
Listen and post your considerations...
I have now tested your VBR mode on thewayitis.wav, fatboy.wav and short.wav (and compared it to LAME --r3mix / --dm-preset standard, using win32LAME 1.7). No ABX, just a quick-n-dirty test, but the difference was obvious enough.
It easily beats both LAME settings on short and fatboy, on thewayitis they all sounded transparent to me. I could still hear a difference to the original on fatboy and short, but both r3mix and dm-standard were so much worse. (Sorry Dibrom, I haven't tried your new settings yet, I don't know where to get my hands on a compile that includes them, can anybody tell me?)
I'm amazed - lower bitrates than r3mix (on normal music, that is - on those samples, the bitrate absolutely exploded), and FAR better quality. I think this has finally convinced me to switch over to OGG - hats off to you, Garf, this is just too great for words! (Take that from somebody who - until now - was 100% sure that --r3mix was the perfect quality/size compromise...)
And the 128k mode is also well tuned, is it? Fine, that might be enough to convince some friends to use it instead of LAME 128kbps without nspsytune...
The only thing that bothers me is the OGG tagging system, which I seriously dislike... firstly because of its structure and secondly because there are no batch taggers around that can handle it, but I guess there's a solution to that somewhere.
CU
A happy Dominic
Originally posted by Volcano
And the 128k mode is also well tuned, is it? Fine, that might be enough to convince some friends to use it instead of LAME 128kbps without nspsytune...
Well, no, it's not really tuned. The standard RC2 128kbps mode was 'ok' for me if it weren't for the stereo hiss. The tuned 128kbps just removes the stereo hiss, at the cost of some more bits, so it may average at a bit more than 128kbps (but it also contains a fix which brings bitrate down again, so maybe it'll even out). I think it will do badly on preecho-heavy samples, perhaps worse than a well-tuned MP3 encoder. For normal music, it should be pretty competitive.
The 128kbps mode that is in current CVS may do better than the 128kbps mode in my encoder.
The only thing that bothers me is the OGG tagging system, which I seriously dislike... firstly because of its structure
Not sure what you mean here...could you explain?
secondly because there are no batch taggers around that can handle it, but I guess there's a solution to that somewhere.
Most people on the vorbis lists use scripts to handle the tagging. I don't use tags myself.
Thanks for your feedback. I appreciate it.
--
GCP
There is one mass tagger for ogg files that I know of.
http://www.mediajukebox.com (http://www.mediajukebox.com)
Bit unstable, and not that easy to use, but it pretty much does the job.
gnoshi
btw... can't wait for rc3 =)
good work garf
Media Jukebox does it? Cool, I'll try it when I get home.
Does it also batch-tag files according to the file name scheme?
CU
Dominic
Problem solved - the answer is "Yes, it does batch-tagging". Works like a charm! I'm beginning to like Media Jukebox, although it's a bit bloated in some places...
Thanks, gnoshi, for the hint!
CU
Dominic
I love the tagging system. It is supa-flexable! I did some audio captures of some of the funny dialog from the tick and it works great. With MP3 IDVx you just have to live with and work around the built in tagging sys. With Vorbis I can create my own like SHOW=The Tick, EPISODE=3, TITLE=Whatever I want. It could be confusing at first, but it allows for much better tagging. No more hard to read abreviations and trying to figure out what each lable reffers to. It is clear concise and there are even tools to edit them inside Windows Explorer!