I did a bit of search here, but i didn't find much info on encoding to very low bitrates, let alone a listening test. Excuse me if such a thread alredy exists.

Anyway, is there any suitable tool apart from Real Audio and Windows Media?
(I saw a mention on Garf's homepage about extreme low bitrate Vorbis encoding, but the samples are gone. Anyone knows about this project?
BTW for me it seemed that going below -q1.5 is not a clever choise, it's much better to downsample input. I never managed to get enjoyable Vorbis files of bitrate less than 35-40 kbps though.)

I did a little non-blind comparison with RealProducer v10 "basic" and Windows Media Encoder v9.1. The latter did a quite impressive performance, one of the test samples was of surprisingly enjoybale quality at as low as 20 kbps (stereo, 22kHz, double-pass encoding). Real Audio didn't rock on the other hand, it didn't show up with a quality pleasant to ears even at 48 kbps. Real files seemed to have strange artefacts and somehow washed-out, empty tones on all records.
The only reason I found this result strange is that a national radio station broadcasts streaming 20kbps Real (among other formats), and (for me at least) it sounds way better than my home-encoded files. Maybe the professional encoder does a better job?
It would be very hard to admit that a Microsoft pruduct is vastly superior on any computing task any ideas?

(Excuse me for potential grave mistakes... 1st post, no expertise )

Try aoTuV (or faster Lancer) OggVorbis encoder with -q-2.
HE-AAC might be worth looking too.

Thx, I did try that, but my samples suffer heavily even at -q1, especially piano. My English is poor to nicely describe the exact nature of degradation, but believe me... classical piano in Vorbis under qual 1.5 just sucks, especially high and hard hit notes Or hear it for Yourself. Some other classical instruments also get tormented very audibly at qual 1 and lower, but not as badly as piano. It's not only very different from what it should be, but very unpleasant! On my samples, Windows Media @32 rocks compared to Vorbis q0, even though I can hear that hallmark metallic edge and some loss of sound depth distinctly.
I did try to play randomly with the extra settings, but had no luck (ofcourse I do not understand what these settings do )
Situation is rapidly improving while raising qual to -q2; I didn't succeed to ABX some short q2 files (aoTuV b4) from the original --doesn't mean a lot though; presently using Realtek onboard chipset w. Sennheiser HD437 lowend headphones and an untrained pair of ears ;-) by a quite loud computer.

I shall try AAC shortly if the encoder is free to try -- and share my experience.

Edit: oh, just another thing...

This station also broadcasts Windows Media at higher bitrates (64..192 kbps), so they definitely have licensed the format -- no reason not to use WMA @20 instead of Real unless Real is better! Of course this can also be due to compatibility questions or following some tradition, but not having WMA@20 should indicate something... should it not?
(No actual worry about Bartók Rádió Budapest -- they also give 320 kbps mp3, and I happen to have DSL )

i have them, and just moved them to be accessed from http here: http://xmixahlx.dyndns.org/audio/samples/vorbis_floggy/

the project is dead, and as rutra80 mentioned, aotuv @ quality -2 is the lowest you'll get with a vorbis encoder that is currently in development.

garf's floggy encoder (low bitrate) is included in GT3B1, posted here:
http://www.hydrogenaudio.org/forums/index.php?showtopic=6023

you'll need to use --resample 8000 (through 6000) IIRC at qualities -1 and -2


later

- Under 18Kbps AMR-WB+ is unbeatable. But I could not find any player yet, only the encoder and the decoder.
- From 20 to 36Kbps use HE-AAC v2 (AKA aacPlus v2). 3GPP reference encoder available, many players can handle this format.
- Over 36Kbps the best should be HE-AAC (AKA aacPlus), many encoders and players.
- The next one could be Vorbis aoTuV or AAC-LC and the border is not cleary defined.

If you can not find an AMR-WB+ player, then you can still use HE-AAC v2. Both codecs are part of the 3GPP release 6 standard for mobile devices but the last one is also part of the MPEG4 standard. You can listen how HE-AAC, less quality at very low bitrates than HE-AAC v2, performs at 24Kbps here. Use the latest Winamp or Foobar2000.

Regards,
Oki

I found a 32kbps listening test at rjamorim's site, hope it will be useful:

Dial-up bitrate listening test discussion thread

Roberto's public 32kbps listening test results

(Roberto's public listening tests)

edit: added "discussion"

Thanks for Your extensive answer! I'll try them all out

I did an HE-AAC Encode of a song usin the NERO -tape- quality preset.

came out to around 30-40kbps

very listenable ... a couple of my friends even said it sounded as good or better than 128kbps mp3s !

Then i started to paly around w/HE-AAC even more ... the portable HE preset comes out very nice results ... imma try to put them onto my ipod soon to see if i can fit 500+ songs on a 1gb shuffle

Get your friends' ears checked ASAP.



The iPod won't play the HE part in HE AAC.

Roberto, your listening tests are really professional and unbiased since you take into account those without a super hi-fi system and an anechoic chamber. I agree with you in that as you said here.

I am afraid that the vast majority of the listeners are like edekba's friends and their subjective perception is relevant for most people out there. I know that saying this statement in HA can be considered a sin but not all the people is an expert artifact hunter with a 120€ headphones like Guruboolez and most of the people listen music under a really bad environment. This is why I prefer your analyses.

I would like to see one of the amazing Guruboolez analyses under not ideal conditions, the result would be then more interesting and relevant for the typical Joe Sixpack.

I also have a question about the 32Kbps related test: Is The Nero encoder you used really creating HE-AAC v2 streams or only HE-AAC?

And a petition for future low and extremely low bitrates listening test (probably this should go to other section of this forum):
1. It could be interesting to see how Vorbis aoTuV performs against AMR-WB+ and the 3GPP HE-AAC v2 encoder.

Thanks,
Oki

lol yeah i know :-/ I can but they are all about bass and stuff they can get from p2p :-X

about the HE-AAC ... Doh :-/ maybe i can just try regular aac @ that low of bitrate ...

thanks for headup tho, before i went crazy w/the HE-aac transcode heh.

It would be interesting, but not necessary "more relevant". A vast majority of people on this board are not really interested by 64 kbps encodings or 128 kbps made by Blade or other crap MP3 encoders (on which are based the famous xxx [put here what you want] at 64 kbps = MP3 at 128 kbps.)

By the way, I fear that SBR based encoders are not really suitable on any headphone, even crap ones. I tried once with 4 € earbuds connected on AC97 - not really an ideal situation... - and most "SBR artefacts" were still audible. By SBR artefact I mean the "grainy" sound and agressive noise you can hear on all SBR encoders: MP3Pro and AAC-HE from Coding Technologies and Nero Digital.
But on speakers, especially poor ones, I don't think that such artefacts are still audible.


Rasputin> Vorbis don't perform very well with several classical recordings at low bitrate - sometiimes worse than WMA indeed. I agree that piano (high dynamic range) suffers much than other instruments.
Try AAC-HE profile. It should be more pleasant at such low bitrates.

Did you try a new HE-AAC2 in winamp 5.1 .
for 9$ earphones . Foobar player, DSP enable (only equalizer) :
1. Lame 128 kbit/s
2. HE-AAC2 Winamp 5.1v 64 kbit/s (very close to lame 128)
3. db HE-AAC2 64 kbit/s (aka enchased) (close to HE-AAC2 Winamp)
4. Nero HE-AAC1 64 kbit/s (far to be close to anything acceptable)

I am quite sure you used player that does not decode HE-AAC data from the MP4 files.



It was HE-AAC v1 - Nero HE-AAC v2 encoder is coming to public soon, and should be pretty much better than the v1

I'm very sure that it was foobar 0.9 beta 7 that decodes well HE-AAC1

These are good news!!!! AMR-WB+ is going to be used a lot in extremely low bitrate applications. Any news about an AMR-WB+ player?

Regards,
Oki

TCPMP( http://www.mabin.info/tcpmp/page.php?id=home&lang=en ) can play AMR-NB(.amr) and AMR-WB(.awb) files!

AMR-WB+ is not the same than AMR-WB, but it is only one little step ahead.

Regards,
Oki

The big difference is the encoded bandwidth -

AMR-NB: 3.4kHz
AMR-WB: 7kHz
AMR-WB+: varies from approx. 7kHz to 19kHz

Just wondering, but did you replaygain the streams? If you didn't, it may be clipping that you hear as "artefact" (at such low bitrates the peaks increase very much because of the aggressive lowpass filtering), so especially music being at the maximum amplitude in uncompressed form already will clip a lot so you have to decrease gain to stop the clipping. This is what I think the radio stations are doing.

Orban is a professional solution for digital radio streaming and broadcasting. They reveal some of their secrets in this document. This is a quote taken from the mentioned paper:
"Experience has shown that a combination of multiband compression and sophisticated peak limiting is the most effective way to broadcast a louder and punchier audio signal."

This is more or less what gasmann was suggesting but there are more elaborated solutions like the one used by XM Satellite Radio. Take a look at this system from Neural Audio, it is used by XM Satellite radio. It claims a 30% improvement in quality when used as a pre-processor before encoding.

Regards,
Oki