Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Converting FLAC to Opus. 160 or 192kbps for universal transparency? (Read 15669 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #25
There is no universal transparency with any lossy codec, regardless of bitrate. All lossy codecs have killer samples.
I agree, but I think it becomes an issue when a codec jumps from "imperceptible" straight into "slightly annoying" area, skipping the "perceptible but not annoying". This is an issue with Opus to me personally. On a contrary, Musepack is noisy at HF and I can ABX it on some tracks, but that noise doesn't annoy me as much as smeared cymbals of MDCT codecs at low bitrates that Opus sadly produces (not always of course, it's still a VERY robust codec after all).
Opus VBR 256 + SoX

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #26
If opus was tuned more for high bitrate, You would have a not annoying result at 192
and perfect or close at 224.  You shouldn't normally need 256 or 320. its IMO because
something somewhere down the line went wrong. Exactly which encoders or versions
I cannot say for sure. But I remember way back lame 3.90 APS was hard, as was musepack.
Now nearly everyone can abx so many things on new codecs @ 192 or more. I am more
confident with mp3 in a way. Why even use opus 192 over lame 3.100 -V 1 then ?? Or even
the older lame 3.90 APS , APX.  You can even try lame 3.100 -V2 --vbr-old.
Who would produce positive easy abx for - V1  on stuff like metallica load ? No one
I guess and even then others won't be able to hear it.
wavpack -b2.9cs.5
lame --preset cd -f

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #27
If opus was tuned more for high bitrate, You would have a not annoying result at 192
and perfect or close at 224.
You seem to be extremely right about this. I have tried Opus 128 a little more, and the result with some of the genres are the same I mentioned before, but cymbals in Metallica - Load are not worse at least and actually even better to the point of transparency for me at most. So tuning is a big thing here, Opus seems to be more consistent at 128 kbps than it is at 192 kbps.

Which is a sad thing because Opus at 128 kbps is not transparent enough for me, as I said earlier regarding Eurobeat genre.
Opus VBR 256 + SoX

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #28
Would --cvbr  or even --hard-cbr  help ?
wavpack -b2.9cs.5
lame --preset cd -f

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #29
Maybe... CVBR gave me good results on Within Temptation - The Cross when I wasn't critical listening. Need more testing.
Opus VBR 256 + SoX

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #30
I did more listening. Generally yes, 128 kbps seems to be the most tuned bitrate for Opus (NOT the best sounding but the most tuned). If it fails to provide transparency, it fails to do so in CBR, CVBR and VBR mode. I can also say that CBR mode is clearly worse than CVBR and VBR to my hearing. If we compare CVBR to VBR, I'm not sure which one is better. VBR is very good in Opus.

For killer samples. I have noticed another killer sample in Papa Roach - Infest, and yes it's a hi-hat again in one specific place that draws attention while listening. I was only able to make it "noticeable but not annoying" at VBR 192 kbps (I could still ABX it), and I achieved full transparency for my hearing only at VBR 256 kbps. Headphones are AKG K612 just in case.

Musepack seems to handle cymbals a lot better than Opus to my hearing, because it achieves full transparency there on lower bitrates. Q6 is transparent in everything I have tried (yet), and it's something like 200 kbps.
Opus VBR 256 + SoX

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #31
.
Opus VBR 256 + SoX

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #32
Are you saying that beyond ~128k, codecs other than Opus are better? even MP3? Or just "safer" in the sense of less likely to produce audible artifacts?


Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #33
Today I have made an interesting founding: I had opusenc 1.3.1 of some unknown subversion before, today I have updated it to 1.3.1-158 and noticed that while some artifacts are still perceptible on samples I named before, they sound different: they have moved into "not annoying" area for me and became harder to ABX. I don't know why, maybe different compilers matter as well.

So I'm not sure what to answer to your question now actually. Previously - yes, I would say that QAAC and Musepack are safer on higher bitrates than Opus, but not now. I'm confused a little bit :)
Opus VBR 256 + SoX

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #34
Love how every thread just turns into a fundamentals discussion ^^
Are you saying that beyond ~128k, codecs other than Opus are better? even MP3? Or just "safer" in the sense of less likely to produce audible artifacts?
I have seen a listening tests that suggest that opus falls flat of aac over 128 kbps .

It would make sense imo, since opus does force you to resample because it needs to work well on mobile devices and doesnt let you use the added bitrate for frequencies over 20khz at all.  Testing this theory though is a bit impossible by ear i think.

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #35
since opus does force you to resample because it needs to work well on mobile devices
Can't find a link but I've seen a statement that the main reason to choose mandatory 48 khz was better pre-echo artifacts suppression. Thus it should be better at high bitrates too. Leaving devices out of their real-time resampling work is just a nice bonus but not a goal.

and doesnt let you use the added bitrate for frequencies over 20khz at all.
Definitely not a reason. Actually that's the part where Opus devs were 100% right with their decision.

Talking about tests - IgorC proved the opposite in his ABC/HR test of high bitrate VBR codecs. Opus was the only codec at 192 kbps that gave him fully imperceptible result. https://hydrogenaud.io/index.php/topic,120007.0.html
Opus VBR 256 + SoX

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #36
Can't find a link […]

The result  is from my test:
https://hydrogenaud.io/index.php/topic,120166.0.html

OPUS doesn't outperform AAC on this test but keep in mind that this result is based on classical music samples only. I planned a second test with different samples at the same bitrate but I had to cancel it.

I noticed on other listening tests of mine (you can browse different the “listening test” section of this board to find more information) that OPUS usually performs much better with pop/rock samples, or modern music samples, than with classical music at a given bitrate. It was really obvious in the 48…64 kbps range. I don't know why: because classical music is usually less loud, or because OPUS struggles a bit with tonal samples? Harmonics? Lack of tuning? Tuning emphasized on modern music and speech?
Anyway I really insist on the fact that OPUS isn't bad with classical music in my opinion, it's just not as good as expected for a modern (>2010) format. But the OPUS format offers a great efficiency with modern music especially in the 60…90 kbps range, and is probably on top at 128 kbps.

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #37
I don't know why: because classical music is usually less loud, or because OPUS struggles a bit with tonal samples? Harmonics? Lack of tuning? Tuning emphasized on modern music and speech?
Let me have a guess - Opus struggles with classical because it has a hardcoded 20 kHz passband, so lower pitched sounds get less bitrate than they would with AAC, which has dynamic highpass AND it's always lower than 20 kHz. But that's just my guess.

P.S. To clear the misunderstanding - I was talking about prooflink for a statement given by me above (about 48 khz letting Opus handle pre-echo better). But it was great that you shared your thoughts and some details about your test. By the way, the test is phenomenal, I really appreciate your hard work!
Opus VBR 256 + SoX

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #38
I don't know why: because classical music is usually less loud, or because OPUS struggles a bit with tonal samples? Harmonics? Lack of tuning? Tuning emphasized on modern music and speech?
Let me have a guess - Opus struggles with classical because it has a hardcoded 20 kHz passband, so lower pitched sounds get less bitrate than they would with AAC, which has dynamic highpass AND it's always lower than 20 kHz. But that's just my guess.

Noticed there no data on what the long block & Short block sample size are. I've got some ambient/electronic albums where I get MP3 style bite rate bloat seen in V2 ~ V0, Like any other codec does It at 48 ~ 96kbps but with Opus at 128kbps It pumps out a 180kbps file. I might be super wrong but Opus seems act like It long blocks are at 576 while short blocks are 120, It trying keep 20KHz no matter what seems to give It sfb21 style issues that MP3 suffers from.

So It not just Classical music that has issues.

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #39
Noticed there no data on what the long block & Short block sample size are.
Figures, because Opus blocks are fixed size (20 ms in my case).

I've got some ambient/electronic albums where I get MP3 style bite rate bloat seen in V2 ~ V0
Figures too, Opus bumps bitrate on transients, so it may take some squarewave-like synths for a series of transients, which happens more often in electronic music than in other genres. That's exactly the reason why Opus is great for electronic music - no pre echo issues, great transients.
Opus VBR 256 + SoX

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #40
Figures, because Opus blocks are fixed size (20 ms in my case).
Opus (CELT) has 2.5ms "short" blocks, but Opus also has per-band time/frequency resolution switching, so the effective block length may be longer or shorter for each individual band.

Figures too, Opus bumps bitrate on transients, so it may take some squarewave-like synths for a series of transients, which happens more often in electronic music than in other genres.
I think it's more likely that Opus will see the square wave's strong harmonics and boost the bitrate to more accurately reproduce them. With 20ms frames, there are only 50 frames per second, meaning the frequency doesn't have to be especially high to detect it as a tone instead of a transient.

From what I understand (and this may not be entirely correct), Opus (CELT) relies on a clever psychoacoustic trick: if a particular frequency band contains noise, it can be stored with less precision because the quantization noise tends to sound similar to the original noise. Some synths produce very little noise, which means Opus can't cut the precision in any bands without reducing the apparent quality, which means Opus will need more bits than usual.


 

Re: Converting FLAC to Opus. 160 or 192kbps for universal transparency?

Reply #42
The problem at the end of the day is that opus is not mp3 or aac and
not lossless. Yes I agree that 160 should be very good and more is the start
of a slippery slope.  However, Its could be better to have a slope with mp3 IMO.
256 mp3 is not inefficient if you factor quality + total universal support ( perfect with cbr 256 or more).
Its also huge saving over lossless in most cases .
wavpack -b2.9cs.5
lame --preset cd -f