Hi,

I just started transcoding my FLAC albums (I'll save them) to Ogg. I first tested q 5, q 6 and then q 4 and I can't tell any difference between them.

But I cant believe that 128kbit files are absolutely transparent? I know it's very subjective, but I wanted to ask you if it's also transparent to you or if you hear any artifacts?

Greets

Edit: I use AoTuV beta 5

Yes, especially with aoTuV b5, it is transparent to me. I guess even lower q might be transparent (I just barely ABXed q1 right without good headphones and just one sampe), but then again... I know what you are talking about. We need to get out of this 7 year old mindset that 128k BLADE mp3s sound crappy (they didn't, to me, back then (computer speakers) - but now they do =)). Modern encoders really rock, and it is not 'god given' that 128k=crap.

I am ripping my old CDs right now, and for most pop samplers and electronica q4 should be absolutely sufficient. For my more 'valuable' and metal CDs I use q5 however... not that it would make much of a (audible) difference, if any at all, but here I want a little greater 'safety margin'.

In terms of frequency, I've still (bleh, I'm 22 ) got a rather good hearing (~19kHz), and that is just the lowpass at q4 - wonderful.

BTW: with the new b5, I can see (by no way hear) low-level noise in the spectrum at about 20kHz (with 1kHz bandwidth) when using q5 that did not appear when encoding to q5 with aoTuV r1. What is the cause of this?

-q2 is transparent to me in casual listening. Maybe I could ABX it, but it sure sounds great to me either way.

-q3 was too hard for me to abx, so I didn't try to test -q4 ^^ at all.

I use Ogg Vorbis aoTuV -q 3.5 (estimated 120 kbps) to encode my music for portable listening on my Cowon iAudio G3. I did not notice any annoying artefact till now, quality is amazingly good for such a bitrate.

That makes me worry a bit.. Is the r1 a stable final release? Should I better use this version than the beta 5?

At q4, I can ABX some rare tracks. I failed all tracks proposed in multiformat ABC/HR tests.
At q5, I can ABX 3 killer samples.
At q6, I can ABX 1 killer sample.
At q7, everything is transparent to my ears.

r1 is a stable release (rebranded b4.51), yes. b5 is IIRC mainly tuned for lower bitrate improvement. However, as I stated, bandlimited, low-level noise at 20 kHz cannot be heard, so don't worry - I'll rather have this AND the newest tunings. I just wondered why this was there (noise shaping? new algorithm?), as this only seems to appear with q5 and not with q4.

When I start using vorbis, in 2002 (v1.0), I encode my files using -q2, because it sound fine to me... then I swith to -q4 because I ABX -q2 and found that is hard but posible that I hear artifact and because the listening test showing that vorbis at 128kbps is just great... but I swith mostly because the placebo effect that if I hardly hear artifacts at -q2, then -q4 will be the perfect balance.

I made a test to abx 128kbps and found that to my ears is almost imposible to tell the diference. All of this using old encoders (maybe aotuv b2), and with AoTuV Beta 5 this will be just the best choice for a nice balance.

In synthesis, -q4 for me sound just perfect. I posibly can ABX it and hear some diferences but... if I only listen music in my computer then I hardly can said that there is a diference.

In plain English.... yes is transparent to me.

Tip from experience: Don't worry about transparency. Worry about music quality. Especially since you have the FLACs.

I encode, and transcode (aaaaah! the horror! oh humanity!) my music collection to -q 1 using aoTuV b5... and I never regret it. Except for classical music. guruboolez scared me enough to do them at -q 3.

Annnd if you see the music in my PDA... you may scream outright: None is higher than -q 0. Most (i.e. nearly all) is -q -0.5, and some even made it to -q -0.75 ... and these with aoTuV R1 (too lazy to re-transcode using aoTuV b5 )

I recently encoded my complete CD collection to FLAC 1.1.3 beta 2 and aoTuV beta 5 at -q 2. I've been making a few ABX tests shortly after the beta 5's release, and I must admit that I already had a troublesome time ABXing many -q 0 samples, causing this bitrate to have become the one I use for my flash-based portable player. At -q 1 I entirely failed distinguishing the Vorbis files from the FLAC ones; since I tested only a few samples I moved one more quality step up to make sure that the whole audio collection sounds transparent to me. Hence I can't even see any need to use the -q 4 setting you asked for. Of course I'm talking about my own hearing here, you don't have to agree with my statement that -q 2 would be transparent. With this subjective claim I'm risking being snubbed due to TOS #8 anyway, because I neither have the ABX logs anymore nor do I have the nerve to do a test at the moment.



Well, we're talking about a modern codec here, therefore you shouldn't compare its bitrates to the ones that are usually used to reach transparency with the good old MP3 format. If we kept encoding to the same bitrates as we always did there wouldn't be any reason to use anything besides LAME, since it features the best possible compatibility and even many modern codecs' features, like VBR encoding and gapless playback.

Edit: Small addendum.

I don't encode to a certain level, it depends on the CD or source really. The highest i'll go is-q 6 for the fact of the lossless stereo coupling. The lowest i'll go is -q 4. I'm not saying I can ABX below -q 4, it's just I use these files on a lot of different systems (car, 5.1 amp, iPod, etc) and like the safety margin of having at least ~128kbps tracks. Even so, i'm probably being overkill.

Small Edit...

For me it is still hard to encode something lower than q6, even when I'm aware that q2 in most cases is audibly the same...

Did you read about loseless stereo coupling at q6 or why do you choose that quality?
To me, reliable quality means 1 quality value above the setting needed for transparency.
(transparency based on a few test samples)
I hope you never run low on space with this setting.
Not to be able to choose q3 because my portable player only has 512MB
is painfull

-q2 is transparent to me in casual listening (home stereo, portable and computer).
When I listen to music I don't try to find flaws all the time like some paranoid freak. IMO on casual listening you don't need to.
I may even go lower to -q1 or -q0 in the future.

I encode at -q6, because the second sample I ever encoded with Vorbis happened to be one of the rather rare Vorbis problem samples where the trouble is very audible. The problem went away at -q6 (i.e. probably with lossless stereo coupling), so I chose -q6 as my standard encoding level. Plus, my audio collection grows slower than my disk space, so I'm not worried about the size.

When testing normal rock/pop samples, -q2 is usually transparent to me.

I switched to HE-AAC for DAP/mobile listening, but -q2 was transparent to me.

-q1 is transparent enough to me.

Look at the results of the Public Multiformat Listening test at 128kbps made by HA members. For many people, in many samples, vorbis was transparent. The low-anchor is probabily what you was expecting for a codec in 128kbps. I was the Anon26.

My experiences with older versions of AoTuV vorbis:

In -q4 I can barely hear the diference in some samples when comparing with the original. That, or -q6, is my choice for transparent encoding, depending of how much space I want to spare. When I want very high quality in less space, I use -q2, where I can hear some artfacts in many samples, but nothing anoying, especialy when not comparing with the original.

For simply listenable quality, I go with AACv1/v2 at 64~32kbps.

It's good enough for me usually... when I'm just listening to music... not audio quality. It'll be fine for you.

I ABX'd tons of songs, and found that for the most part, Q2 is transparent, but there were a couple Q3 songs (very rare). I use Q4 because I know it's overkill and there's no way I could ever tell the difference

I ABX'd -q1, -q2 and -q4. I basically concluded that for a portable flash player, -q1 is the best bang for the buck.

So...I use Squeezebox and FLAC for my home stereo and Vorbis -q1 for muisc on the go.

I have just decided today to delete the q5's from my ipod and encode to q4 today. I could abx q1 and q2. q3 was HARD for me to ABX, but I eventually found Radiohead's "Subterranean Homesick Alien" easy to ABX because of a nosie normalization artifact. q4 doesn't use noise normalization though. I tried to ABX q4 once and gave up because I knew I couldn't do it with my equipment.

But yeah, 128 MP3 and 128 Vorbis aren't comparible; Vorbis is newer and more advanced/complicated. If YOU can't hear the difference, then don't waste the space.

That's about how it is for me. Even when I can tell the difference, it's usually small. At Q7 I doubt it is physically possible for a human to hear the difference. Maybe in some rare cases. For casual listening, Q4 is good enough.

I used to use Lame -V 2 --vbr-new. Am perfectly happy with q4 AoTuV. I still shake my head and marvel at how good tracks with an average bitrate of 125k or so sound.

I'm -q1 all the way. I use FLAC for my home stereo setup (Sonos) and Vorbis -q1 for my flash based DAP...allows me to carry around ~35% more tracks vs -q4 with no discernible difference (to me).

My recommendation -- setup some listening tests and see how low you can go!!

I use aoTuV-b5/Lancer encoded Vorbis track files at it's default compression level(-q3, which averages at about 112KB) for PC playback(and FLAC images for archiving), since in ABX tests then i personally find -q3 to be perfectly transparent to me. I didn't try to go lower though, as even though i maybe could get away with using a lower setting, then i thought that it would be wise to have a little safety margin for possible problem samples. Before changing to Vorbis, then i used LAME at it's -V5 setting and so by using Vorbis instead of MP3, then i am able to save a little extra space on my lossy collection, since Vorbis -q3 uses about the same average bitrate as LAME -V6 does

With aotuvb4 to b5, -q4 makes my music is transparent to me about 80-90% of the time.

I haven't tested older versions, but i remember lots of bad noises in -q4 w/ libvorbis 1.0 back when I first really got my own computer.

I've abxed quite a few tracks, and most the time aotuv is transparent to me, only a few choice tracks have not been. Comparing it to LAME and iTunes AAC, it always gave me a larger quantity of transparent results, so I prefer it.

Since my iPod is now broken, I'll be looking into a new player that either runs rockbox, or supports Vorbis natively.

I stick with -q4 for size reasons. -q5 would probably make everything I have transparent, but then again I can't drag myself to go lower then 128kbps either to save more space. It's an OCD thing.

Are you referring to the occasional 'stuff' above the lowpass frequency (as seen with a spectrum analyzer like analfreq)? That has been happening with all versions I have tested. Drop the lowpass to 10khz or something and see if you can hear it (try using wavgain before encoding and see what happens )

Is q4 transparent for me? In general yes, but I don't feel like wasting that much space when q0 with the lowpass boosted makes me plenty happy.

I did a pratical listening test to figure the superiority of OGG over MP3.

Depeche Mode - Policy of Truth (Violator, 1990)
Kraftwerk - Homecomputer (Minimum-Maximum CD2, 2004)

These two songs have quite a good dose of well mastered bass, and loads of cymbals and subtleties.

Each song had a version:

WAV
OGG Q2
OGG Q1
OGG Q0
LAME V6
LAME V7

With LAME, you can easily pick up that V7 is unusable, coz it drops straight to 32kHz in that setting. V6 would be pretty transparent if not for the bass that is less present and less dense than the WAV.

With OGG, the funny thing is that all are seem transparent and similar to WAV, I kept encoded all the way down to Q -1.0 to see if it became anything like LAME V6. And no... not even the Q -1.0 setting was anything like V6, but with much superior quality. Although the Q -1.0 I found the cymbals a bit "fuzzy" compared to WAV. But it was quite transparent at Q1, Q0.

Of course, these are my personal testing, for what I could catch up. I ended up with the files:

Policy of Truth WAV - 50MB
Policy of Truth V6 - 4.4 MB
Policy of Truth Q0 - 2.3 MB

Homecomputer WAV - 61MB
Homecomputer V6 - 5.7 MB
Homecomputer Q0 - 3.0 MB

With OGG artifacts are just a bit different than mp3 - coarse fat sound, noise, stereo image irregularities. I can pick it up at 64-80k even outside abx. I doesn't sound half bad for the bitrate and aotuv version are much improved, but I would not want to listen to it at home on music that I like.

Offtopic

Nice avatar shadowking
Paradise Lost are great.

Hehe.. Thanks , I want a PL tatoo !

I "ABX"ed with Winamp a while back. I created samples and put them in the Winamp playlist in random order, so I don't know if this counts as "ABX".

Result: I was always able to ABX -q1, barely able -q2 and -q3 was transparent to me 100% of the time.

At -q2, the only samples I was able to ABX were all of the psy trance samples (the beat sounded flat) and AC/DC's bell in the into of "Hell's Bells".

I encode at -q5 because of plain paranoia


Edit:
Paradise Lost rules Draconian Times is still one of my favorite albums, even if I bought it right after it got released.

I can ABX q2, but i'm fine with q1 for portable listening. At home i got my .flac files.
As for 128kbps, remember that it scores very high on listening tests even with LAME MP3. IMHO the idea that more than 128kbps is somehow required for decent quality is horribly outdated.

Anyway, this thread needs a poll. It'd be useful to see what -q levels are most common these days.

I agree. I suggest multiple questions:
- What -q level you use at home (with option: No Vorbis @ home)
- What -q level you use on the road (with option: Never on the road)
- What -q level *in your experience* is transparent enough for at least 90% of your music collection (with option: Never test this objectively)

The -q level should range from -2 to 6 (10 options only, IIRC. Besides, I don't believe -q 6 is not transparent enough for 90% of the musics in the universe)

Any other suggestion? If no, then someone should start the poll.

Edit: Mistyped '6' as '7'. Corrected.

Or perhaps the first & last options should be "less than -q-1" and "more than -q5" or something like that.

The results of the poll (if it gets done) would be interesting, but as has been said many times before, one's own needs/priorities/perceptions are more important for these things than everyone else's; after all, I'm the one listening to my music, not everyone else.

Edit: forgot to say that I use -q2 for my CDs and -q3 for friends' CDs (ahem! - purely for evaluation purposes only! Of course I delete the files after I've 'evaluated' the music, your honour! Each and every time! )

*cough* I usually *cough* forgot *cough* to delete the files...

I have a suggestion, but I'm not sure how to implement it in the poll: try to ascertain whether people got their opinions/findings via ABX or not. Maybe, "I came to my conclusions based on..."?

I haven't been able to ABX anything encoded with aoTuV b5 that's in my music collection at -q4, but I use -q5 as an added bit of security (plenty of room on my DAP). Used to use -q6 but realised I'm just wasting space.

Vorbis has come a long way quality wise thanks to aoyumi. I'm sure there are problem samples out there, but for the majority of music -q4 is impressive (and I'd wager the same for a lot of stuff at -q2 and -q3 though I haven't really tried ABXing much at those -q levels).

Not always. I managed to ABX a few non-killer samples at q5 so I use q6 for my collection.

In finality, I guess the acceptability of -q4 depends on (in no particular order):

- Quality of speakers
- Quality of listeners
- Ambient noise
- Expectancy of listener
- Nature of encoded waveform (i.e. problem samples/killer tracks)

So a listening test is the only way to determine if -q4 (or any value) is transparent for a particular person, with a particular setup, in a particular situation.

I used to use -q6 before switching to MPC for a while. Now that I've back on Vorbis, I'm using -q5.

But I'm still not sure if that's transparent.

With most music, -q4 is transparent to me, but for the safety margin I use -q5 for home listening. On my laptop, where space is quite limited, -q3 sounds good enough for me (for comedy CDs, -q0.5 is sufficient).

pepoluan's (I think) tip to un-train one's artifact listening is really good.

Absolutelly transparent.

Disclaimer: Totally against TOS #8...didn't do much ABXing except occasionally for HA listening test. And I want it that way...I feel no need to train myself in capturing flaws of lossy audio compression.

Yeah it's transparent to me. I could easily abx q0 & q1, I had problems with q2 but I could still abx it. I was unable to abx q3 so I took the safe route and I use q4 for all my music. I tested with AoTuV b5 on my favorite tracks.

I'm sure I can find songs / problem samples where I can abx q4 as well but I don't see the point of it. q4 is simply good enough for me.

/Kef

Yup, that's me:

http://www.hydrogenaudio.org/forums/index....st&p=425910

If one don't unlearn after ABX, his/her ear-brain auditory system will be very sensitive to artifacts, even to the point of hearing artifacts that are *actually* in the original recording. Like I did when I didn't unlearn after some intense ABX session.

If you are not (yet) fluent in unlearning your auditory system, here's a shortcut:

With a DAP, go out to a rather noisy environment (e.g. a park where there's a lot of children playing, preferably near a rather busy road so there's low-frequency rumbling of traffic), and listen to any non-transparent audio track. Lie back, enjoy your time in the shades (keep away from heatstroke, heh), and relax. Enjoy the music. The key: relax. You're unwinding your auditory system.

Then you can start unlearning after you've unwound your auditory system.

Trust me: For personal use and enjoyment, ABX-ing is totally useless. You'll never be able to enjoy your songs again, always worrying that it is non-transparently encoded.

Which is why, in the poll, I did listening tests but not ABX test, to determine what -q I shall use. Just simple test to hear if the song become less enjoyable.

For my mp3 player & headphones, -q2 is sufficiently transparent.

I've been using this encoder:

http://homepage3.nifty.com/blacksword/

under wine in linux, because it seems the default oggenc doesn't support multiple cores, and I like the speed.

Yeah those lancer builds sure are zippy.

Hello!

Same here. Yesterday I spent 1,5 hours trying to ABX ~10-15 samples of
music I like, all of it classical, pop, rock or metal, using foobar2000's ABX
utility. I made samples from flacs and encoded them with -q -2, -1, ..., 6
using libvorbis with the aotuv r1 patch.

Everytime I was able to tell the difference I dropped the rest of the samples
of the same quality. -q -2 to -q 0 were not too difficult, -q 1 and especially -q2
I found hard to ABX. In the end I couldn't ABX any sample I encoded
with -q 3. I had already encoded my collection with -q 4, so I guess I'm
lucky that way, because I'd have added +1 as a safety net.

I'm not deaf yet, I can still hear the the mosquito ringtone loud and clear
and every "broken" tube tv emitting this high pitched noise makes me want
to leave the room. I'm pretty amazed what vorbis and aotuv can do.

Regards
micmac