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: [major Did Not Do the Research] From: IETF Opus codec now ready for te (Read 8100 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Quote from: Gainless on
Well, if you set the encoder to 512 kbps it'll keep the resulting file around that bitrate, no matter how far it could be compressed further losslessly.



Can´t the encoder choose to decrese it but keep it lossless?
Like VBR mode etc?

So if it feel, oh i only need 200 bitrate around here, it will use 200bitrate, even if it says you can use up to 512?

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #1 –
Quote from: zerowalker on
Quote from: Gainless on
Well, if you set the encoder to 512 kbps it'll keep the resulting file around that bitrate, no matter how far it could be compressed further losslessly.



Can´t the encoder choose to decrese it but keep it lossless?
Like VBR mode etc?

So if it feel, oh i only need 200 bitrate around here, it will use 200bitrate, even if it says you can use up to 512?


That calls for a purely quality-based VBR mode, which Opus just doesn't have yet, as far as I know.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #2 –
AFAIK 1.0.1 and experimental branch already have 3 modes: CBR, contrained VBR (CVBR) and unconstrained VBR.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #3 –
I thought the unconstrained VBR still used a target bitrate, rather than some tuned PSNR or something else like that.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #4 –
Unconstrained VBR (UVBR) doesn't keep the target bitrate no matter what metrics do. 

I haven't checked 1.0.1RC but experimental branch doesn't hold the target bitrate on difficult samples at all. Bitrates go well down and high.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #5 –
Yes, but what I meant was, if you set a bitrate that's way too high, it will maintain it, rather than trying to save bits that it shouldn't need to spend to achieve an acceptable quality level. Of course, I guess if you're asking it for 512kbps, it's going to assume you want an insane level of quality.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #6 –
Quote from: kode54 on
Yes, but what I meant was, if you set a bitrate that's way too high, it will maintain it, rather than trying to save bits that it shouldn't need to spend to achieve an acceptable quality level. Of course, I guess if you're asking it for 512kbps, it's going to assume you want an insane level of quality.


But even with that, shouldn´t it NOT spend uneeded bits?

Let´s say that it can be Lossless at 250bitrate (not transparent, lossless).

And you say, Hey i give you 512bitrate to play with;)!

Shouldn´t it use the bits it can use and skip the unseeded?
Not force all the bits cause it can use them?

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #7 –
A constrained mode may force an encoder to fill unused audio frame space with padding junk, to satisfy systems which may rely on constant audio frame sizes. AC3 encoders for DVDs will do so. MP2/3 encoders usually too.

I believe zerowalker would like to know if Opus will pad high CBR/ABR target bitrates beyond saturation as well.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #8 –
Quote from: LigH on
A constrained mode may force an encoder to fill unused audio frame space with padding junk, to satisfy systems which may rely on constant audio frame sizes. AC3 encoders for DVDs will do so. MP2/3 encoders usually too.

I believe zerowalker would like to know if Opus will pad high CBR/ABR target bitrates beyond saturation as well.


Exactly, though i am not talking CBR, but VBR. I am not that advanced in the subject, but i know that VBR will try to use as much or as little as it feels is neccesery withint a certain limit.

Though i myself would personally Love is there was a VBR without a Limit, that would produce Lossless/Transparent sound with the bitrate it feels is needed, through some kind of calculation (PSNR or something, i don´t know?).

Though there i am starting to think about the 512 bitrate settings, which is More than enough to produce Transparent or Lossless sound, so with VBR and that bitrate it should be able to produce it like it wasn´t a limit.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #9 –
If Opus is at all able to work losslessly, then a target as high as 512 kbps is probably enough to get this result. A certain answer has to be given by developers, though...

Subjective transparency will be possible at lower targets already.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #10 –
Quote from: LigH on
If Opus is at all able to work losslessly, then a target as high as 512 kbps is probably enough to get this result. A certain answer has to be given by developers, though...
Subjective transparency will be possible at lower targets already.


Opus cannot do lossless encoding, and if it could 512kbps certainly wouldn't be enough to reach it in all cases (uncompressed 16 bit stereo input can have up to 1.5mbit/s of data). Opus certainly should be able to reach transparency for all samples when encoding at a sufficiently high rate.

There appears to be a large amount of VBR confusion above.  VBR mode makes no promise about the resulting bitrate. The current releases (and git master) tends to land quite close to the requested value on almost all samples, whereas EXP does more aggressive bitrate moving around, and will only achieve the requested average rate on diverse collections.  The encoder has a constrained VBR mode which is more like CBR in high delay formats (varies rate without increasing the worst case buffering delay for a CBR transmission channel at the nominal rate), and a true hard CBR mode which operates exactly at the specified rate on a frame by frame basis and is suitable for preventing information leaks in encrypted channels.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #11 –
Quote from: zerowalker on
Though i myself would personally Love is there was a VBR without a Limit, that would produce Lossless/Transparent sound with the bitrate it feels is needed […] Though there i am starting to think about the 512 bitrate settings, which is More than enough to produce Transparent or Lossless sound, so with VBR and that bitrate it should be able to produce it like it wasn´t a limit.

That’s not how lossless audio works at all. Lossless compression takes the bitrate it needs, without caring about any limit that you might want.

And there is no such thing as “[l]ossless sound”. The closest thing is transparency, but since you’ve included that as an “or” option in the same breath, I don’t know whether you understand its definition either.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #12 –
Quote
That’s not how lossless audio works at all. Lossless compression takes the bitrate it needs, without caring about any limit that you might want.

What he said was 'VBR without a Limit'

Quote
And there is no such thing as “[l]ossless sound”. The closest thing is transparency, but since you’ve included that as an “or” option in the same breath, I don’t know whether you understand its definition either.

I think he does understand. What he meant with lossless sound is that the output bits from decoding the encoded file matches the output bits of the orignial source.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #13 –
Quote from: LithosZA on
Quote
And there is no such thing as “[l]ossless sound”. The closest thing is transparency, but since you’ve included that as an “or” option in the same breath, I don’t know whether you understand its definition either.
I think he does understand. What he meant with lossless sound is that the output bits from decoding the encoded file matches the output bits of the orignial source.
Oh, thanks for that revelation. Yet your pointing out something that is blatantly evident does nothing to alter the fact that “the 512 bitrate settings” of any codec are not “More than enough to produce Transparent or Lossless sound” (i.e. bit-for-bit preservation of the input signal) for the vast majority of material, and so it does not back up that case, rather the opposite if anything.

If zerowalker was simply referring only to transparency, then fair enough, but there was no need to include two non-equivalent terms and therefore introduce ambiguity.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #14 –
Quote from: db1989 on
Quote from: LithosZA on
Quote
And there is no such thing as “[l]ossless sound”. The closest thing is transparency, but since you’ve included that as an “or” option in the same breath, I don’t know whether you understand its definition either.
I think he does understand. What he meant with lossless sound is that the output bits from decoding the encoded file matches the output bits of the orignial source.
Oh, thanks for that revelation. Yet your pointing out something that is blatantly evident does nothing to alter the fact that “the 512 bitrate settings” of any codec are not “More than enough to produce Transparent or Lossless sound” (i.e. bit-for-bit preservation of the input signal) for the vast majority of material, and so it does not back up that case, rather the opposite if anything.

If zerowalker was simply referring only to transparency, then fair enough, but there was no need to include two non-equivalent terms and therefore introduce ambiguity.


It is as he says, i know that Lossless is Bit Exact, not matter if it´s audible or not. Transparent is the "Same" but with removing the Bits that are not Audible (in theory).

I though that some music, speech etc can reach Lossless below 512kbps, but maybe that´s wrong?
I know that the "True" bitrate is 1.5k (Wave), but that´s uncompressed, Flac and TAK can reach far lower depending on the sound, so it can reach below 512kbps for certain stuff, in this particular stuff, i was wondering if OPUS will do the same, Not wasting bits as it will be Lossless at, let´s say 412kbps.

Sorry for not being to technical about it, i am not a developer or Audiophile, just a "wana be" Audiophile or what to call it:)

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #15 –
Quote from: zerowalker on
i was wondering if OPUS will do the same, Not wasting bits as it will be Lossless at, let´s say 412kbps.

Opus is not a lossless codec and cannot be "bit-exact".

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #16 –
Quote from: zerowalker on
I know that the "True" bitrate is 1.5k (Wave), but that´s uncompressed, Flac and TAK can reach far lower depending on the sound, so it can reach below 512kbps for certain stuff, in this particular stuff, i was wondering if OPUS will do the same, Not wasting bits as it will be Lossless at, let´s say 412kbps.


A lossless codec must have a maximum bitrate at least as high as the largest PCM format it can compress.  This is a consequence of how compression works.  However, just because a codec can go up to however many kbps does not mean it can be lossless.  Lossless codecs work differently then lossy.  No bitrate will ever make Opus lossless, and so there is no point in having very high bitrates.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #17 –
Oh, so even if a speech clip is lossless with "200kbit" with FLAC, it´s not "Lossless" in Opus?
Even if it can be Completely Transparent, it will never be Bit-Exact so it can´t be Lossless in terms of the word Lossless (if i get it right?)

Btw, is there any ideas of using an Unlimited VBR to achieve Transparent sound?
I know it´s possible to produce Transparent sound with calculations alone, but it should be possible to do it with some overhead right?

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #18 –
Quote from: zerowalker on
Oh, so even if a speech clip is lossless with "200kbit" with FLAC, it´s not "Lossless" in Opus?


Correct, because Opus is not a lossless format.

Quote from: zerowalker on
Even if it can be Completely Transparent, it will never be Bit-Exact so it can´t be Lossless in terms of the word Lossless (if i get it right?)


Correct.

Quote from: zerowalker on
Btw, is there any ideas of using an Unlimited VBR to achieve Transparent sound?


Yes, in fact this is the entire purpose of VBR.

Quote from: zerowalker on
I know it´s possible to produce Transparent sound with calculations alone, but it should be possible to do it with some overhead right?


I don't know what this means.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #19 –
Quote from: zerowalker on
Oh, so even if a speech clip is lossless with "200kbit" with FLAC, it´s not "Lossless" in Opus?

IIRC lossy codecs like Opus, Vorbis and AAC apply Fourier-related transforms which are lossy.


And if the original souce 48 kHz/24 bits  was encoded by Opus at 512 kbps then where we will put PCM 44.1kHz/16 bits? Lossless or lossy?
44.1/16 is lossless ... comparing to what? Orginal CD format, yes.  48/24 - no. Hence we don't talk about lossless aspect of 44.1/16 anymore but rather that's more than enough (hence transparent). Why we can't do the same for any other lossy encoder? Opus as Vorbis and AAC is higly transparent at 250-300 kbps.
PCM is not lossless representation of analog signal. It's transparent.

Talking about 512 kbps for lossy codecs it's equivalent to talk about 192kHz/32 bits PCM for "lossless".

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #20 –
Quote from: lvqcl on
Opus is not a lossless codec and cannot be "bit-exact".

Quote from: saratoga link=msg=0 date=
Correct, because Opus is not a lossless format.

I'd like to challenge this. Let's say we have an 8-bit PCM waveform and encode it with FLAC and 512-kbps {insert transform codec here}. Then we round the decoded transform-codec version to 8-bit PCM. If you design the transform encoder a certain way, I'm sure the rounded decoding can be bit-identical to the FLAC decoding. If instead of 512 kbps you would use 2 Mbps or so, I'd guess it would also work with 16-bit PCM.

Quote from: IgorC link=msg=0 date=
IIRC lossy codecs like Opus, Vorbis and AAC apply Fourier-related transforms which are lossy.

Nope, the transforms are perfectly invertible, hence lossless. The lossy part of lossy waveform coding is the quantization of the spectrum (and in case of MP3, use of a non-perfect-reconstruction filter bank, but that's a detail).

Chris
If I don't reply to your reply, it means I agree with you.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #21 –
Quote from: C.R.Helmrich on
Quote from: lvqcl on
Opus is not a lossless codec and cannot be "bit-exact".

Quote from: saratoga link=msg=0 date=
Correct, because Opus is not a lossless format.

I'd like to challenge this. Let's say we have an 8-bit PCM waveform and encode it with FLAC and 512-kbps {insert transform codec here}. Then we round the decoded transform-codec version to 8-bit PCM. If you design the transform encoder a certain way, I'm sure the rounded decoding can be bit-identical to the FLAC decoding. If instead of 512 kbps you would use 2 Mbps or so, I'd guess it would also work with 16-bit PCM.


Yeah but if the spec doesn't specify how to handle things like rounding error, its still dependent on you knowing exactly how the specific encoder worked and then designing a decoder around that.  So I doubt that if ffmpeg ever wrote an Opus encoder you could be sure that your decoder would produce bit exact output for all samples. 

Quote from: C.R.Helmrich on
Quote from: IgorC link=msg=0 date=
IIRC lossy codecs like Opus, Vorbis and AAC apply Fourier-related transforms which are lossy.

Nope, the transforms are perfectly invertible, hence lossless. The lossy part of lossy waveform coding is the quantization of the spectrum (and in case of MP3, use of a non-perfect-reconstruction filter bank, but that's a detail).


Well they're invertible when they're done with sufficient precision.  Looking at the opus decoder, theres actually multiple different transforms you can use, and while I haven't tested it, I wouldn't be surprised if they produce different output due to rounding error.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #22 –
Quote from: C.R.Helmrich on
Nope, the transforms are perfectly invertible, hence lossless. The lossy part of lossy waveform coding is the quantization of the spectrum (and in case of MP3, use of a non-perfect-reconstruction filter bank, but that's a detail).

Yes, You're right. Let me rephrase it  :
A transform doesn't bring any compression gain without quantization and zeroing low magnitude coefficients.
 

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #23 –
Quote from: IgorC on
A transform doesn't bring any compression gain without quantization and zeroing low magnitude coefficients.


We don't zero low magnitude coefficients. Sucks to be other formats. 

But there are tradeoffs. Opus really can't do lossless, not at any rate, not even if you substitute the transforms and filters with perfect reconstruction integer alternatives and carefully controlled rounding.

[major Did Not Do the Research] From: IETF Opus codec now ready for te

Reply #24 –
Quote from: IgorC on
Quote from: zerowalker on
Oh, so even if a speech clip is lossless with "200kbit" with FLAC, it´s not "Lossless" in Opus?

IIRC lossy codecs like Opus, Vorbis and AAC apply Fourier-related transforms which are lossy.


And if the original souce 48 kHz/24 bits  was encoded by Opus at 512 kbps then where we will put PCM 44.1kHz/16 bits? Lossless or lossy?
44.1/16 is lossless ... comparing to what? Orginal CD format, yes.  48/24 - no. Hence we don't talk about lossless aspect of 44.1/16 anymore but rather that's more than enough (hence transparent). Why we can't do the same for any other lossy encoder? Opus as Vorbis and AAC is higly transparent at 250-300 kbps.
PCM is not lossless representation of analog signal. It's transparent.

Talking about 512 kbps for lossy codecs it's equivalent to talk about 192kHz/32 bits PCM for "lossless".



Okay i think i get the grip of it:)

But is 250-300kbps really the "Transparency Bitrate"?
I thought Opus was around 152-196 (if you skip the bugs it that´s been discovered).
As AAC is Transparent around 196 i think (Of course it´s not always, but overall or what to call it).


And saratoga, i though VBR needed to have a limit?
it doesn´t calculate what it needs to be Transparent through a calculation alone, it just go with a limit and try it´s best to use what´s needed around that limit (i think?).

What i want is a VBR mode that will achieve Transparency alone, without any bitrate indication, it will just try it´s best to achieve transparency.

Don´t know if that´s possible though, but it´s just a thing i would like:)

And that PCM isn´t Lossless of analogue, do you mean that it isn´t Perfectly Calculated the analogue signal?
As i don´t think digital can make an Exact Replica of Analogue ( i think it´s pretty much impossible as analogue isn´t limited to herz,bits etc right?).