Skip to main content

Notice

If you are using a Hotmail or Outlook email address, please change it now, as Microsoft is rejecting all email from our service outright.
Topic: Classical/opera 192kHz/24bit to Opus (Read 2079 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Classical/opera 192kHz/24bit to Opus

Hi,

I have some classical/opera music 192kHz, 24bit, flac I want to convert to lossy to save storage.

The only codec I found that still supports 24 bit/96kHz is WMA pro 10, which is quite old but sounds right.

I used dbpoweramp 17 to convert from flac, with DSP resampler SSRC 48kHz, then triangular dither 16bit then latest opus encoding at 160 kb/s. I didn't noticed artifacts on both WMA/Opus.

Is there any better lossy scheme to encode my collection?

Best,
Luciano


Re: Classical/opera 192kHz/24bit to Opus

Reply #2
Most of lossy encoders support 24-32 bits.  There is no need for 16 bits.

Also Opus has built-in resampler. So You shouldn't  do anything. Opus will handle 96 kHz/24 bits without issue.

P.S. And as m14u indicates, before going with lossy format, it's a good idea to store 44.1/16 FLAC lossless.  99.9% of people can't hear the difference between 44.1/16 and 96/24.

Re: Classical/opera 192kHz/24bit to Opus

Reply #3
Hi,

I have some classical/opera music 192kHz, 24bit, flac I want to convert to lossy to save storage.

The only codec I found that still supports 24 bit/96kHz is WMA pro 10

I just tried with a 192/24 file of mine, and Ogg Vorbis is able to keep the 192 KHz original sampling rate. Opus, iTunes AAC (qaac), FDKAAC and FHG AAC failed and output seems to be limited at 48 KHz.
But I suggest you to keep going on resample to something more basic like 48 KHz: you can use almost every encoder and format with this one. At 160 kbps, you can use Opus, Vorbis, LC-AAC: they're all really fine.

Re: Classical/opera 192kHz/24bit to Opus

Reply #4
Thank you guys,

So, I will keep my actual scheme :
1) DSP: 192kHz --> 48kHz
2) DSP: triangular dither 24--> 16 bit
3) Opus 160 kbit/s
It's straightforward in dbpoweramp.

Quote
Most of lossy encoders support 24-32 bits.  There is no need for 16 bits.

Do you mean that using dithering is useless here?

Re: Classical/opera 192kHz/24bit to Opus

Reply #5
Yes, useless unless you are using some encoder that requires 16 bit.


Re: Classical/opera 192kHz/24bit to Opus

Reply #7
Keeping the 24 bit depth would be a good idea — it helps to keep the integrity of the audible frequency range (in case you want to raise volume during playback and not hear noise floor), and formats like Opus can handle 24 bits without needing to dither.

I just tried with a 192/24 file of mine, and Ogg Vorbis is able to keep the 192 KHz original sampling rate. Opus, iTunes AAC (qaac), FDKAAC and FHG AAC failed and output seems to be limited at 48 KHz.
Vorbis does indeed support up to 192KHz natively.
AAC does support 96K, just not Apple—probably due to the encoder's 320 kbps limit. Those that do support 96K are: Nero (cuts off at 30KHz); FhG; and FAAC (only useful for high bit rates). It's best to resample to 96KHz before using an AAC encoder if wishing to encode at that sampling rate.
Opus only does 48K fullband and cuts off at 20KHz.

For 160 kbps, it's still better to go with Opus.

Re: Classical/opera 192kHz/24bit to Opus

Reply #8
Thank you, for the 24 bit precision, found this link also confirming the 24 bit input support: 24 bit support

So, the only remaining step would be resampling:

Resampling XiphWiki

... which is also probably useless too.

Conclusion: I will feed my flac 24 bit directly into Opus. I am fine with 48KHz output.

Re: Classical/opera 192kHz/24bit to Opus

Reply #9
Don't dither.    Dither is supposed  to help when you reduce the bit depth but with lossy compression you're not storing the individual samples so there is no traditional bit depth.

...And, at normal-reasonable conditions at 16 bits or better you can't hear dither (or the effects of dither) anyway.

Re: Classical/opera 192kHz/24bit to Opus

Reply #10
Keeping the 24 bit depth would be a good idea — it helps to keep the integrity of the audible frequency range (in case you want to raise volume during playback and not hear noise floor), and formats like Opus can handle 24 bits without needing to dither.
...
Vorbis does indeed support up to 192KHz natively.
AAC does support 96K, just not Apple—probably due to the encoder's 320 kbps limit. Those that do support 96K are: Nero (cuts off at 30KHz); FhG; and FAAC (only useful for high bit rates). It's best to resample to 96KHz before using an AAC encoder if wishing to encode at that sampling rate.
Opus only does 48K fullband and cuts off at 20KHz.

Opus can keep more than 24 bits of dynamic range so no issue here. That's sometimes useful especially for audio that hasn't been properly normalized/mastered yet. When it comes to the sampling rate, it's a lossy codec. Lossy codecs save bits by discarding the details that are least likely to be audible. It means that any extra bit you have should be spent below 20 kHz where you can hear it. And if there's no bits left above 20 kHz, then there's nothing to code there anymore so there's no need for more than 48 kHz sampling rate.

Re: Classical/opera 192kHz/24bit to Opus

Reply #11
Well I make test with FLAC, 24 bits, 96 khz, 5.1 input:

- AAC:

Nero: 96 khz, 5.1, 24 bit input at 2000 kbps max in LC mode
Quote
neroAacEnc.exe -if output-24b.wav -of test-lc.aac -q 1.00 -lc

fdk_aac: 96 khz, 5.1, 24 bit input at 3000 kbps max in LC and 1500 kbps max in HE mode
Quote
ffmpeg -i 2L-084_mch-96kHz_11.flac -vn -codec:a libfdk_aac -profile:a aac_low -vbr 5 -channels 6 output-lc.mp4
ffmpeg -i 2L-084_mch-96kHz_11.flac -vn -codec:a libfdk_aac -profile:a aac_he -vbr 5 -channels 6 output-he.mp4
ffmpeg -i 2L-084_mch-96kHz_11.flac -vn -codec:a libfdk_aac -profile:a aac_low -b:a 3000000 -channels 6 output-lc.mp4
ffmpeg -i 2L-084_mch-96kHz_11.flac -vn -codec:a libfdk_aac -profile:a aac_he -b:a 1500000 -channels 6 output-he.mp4

Apple AAC: 48 khz, 5.1, 24 bits input at 768 kbps in LC mode and 640 kbps in HE mode
Quote
qaac64.exe -i output-24b.wav -o azerty.aac --quality 2 --tvbr 127 --verbose

Nero seem the best with no cut off at extreme high bitrate
You can make amazing AAC file at 96000 khz with SBR (HE-AAC) with fdk_aac at 1500 kbps

- OGG Vorbis

libvorbis: 192 khz, 5.1, 24 bits input at 1500 kbps
Quote
ffmpeg -i 2L-084_mch-96kHz_11.flac -codec:a libvorbis -q:a 10 -cutoff 0 vorbis.oga

Ogg Vorbis seem really simple solution for HiRes quality audio, with really high bitrate and no cut off.

Re: Classical/opera 192kHz/24bit to Opus

Reply #12
Opus can keep more than 24 bits of dynamic range so no issue here. That's sometimes useful especially for audio that hasn't been properly normalized/mastered yet. When it comes to the sampling rate, it's a lossy codec. Lossy codecs save bits by discarding the details that are least likely to be audible. It means that any extra bit you have should be spent below 20 kHz where you can hear it. And if there's no bits left above 20 kHz, then there's nothing to code there anymore so there's no need for more than 48 kHz sampling rate.

In my opinion Opus uses a digital filter that moves the energy present after the cutoff frequency to the previous high frequencies; it is a problem that depends on the digital low-pass filter chosen for Opus and not strictly on the Opus encoder. For this reason, it does not convince me when I compress songs that I have at a sampling frequency higher than necessary.

However, to lower the sampling frequency, it is necessary to use a low pass filter, so apparently it might be better to use a different encoder. In fact, we always have files obtained by resampling the result of the manipulations performed at larger samplings and whenever the sampling or bit depth of a song is reduced, noise is inserted.



MOD edit: (quote)

Re: Classical/opera 192kHz/24bit to Opus

Reply #13
In fact, we always have files obtained by resampling the result of the manipulations performed at larger samplings and whenever the sampling or bit depth of a song is reduced, noise is inserted.
The noise caused by resampling is insignificant compared to the noise caused by lossy compression. If you're worried about the resampler, maybe you shouldn't be using a lossy codec like Opus.

Re: Classical/opera 192kHz/24bit to Opus

Reply #14
In my opinion Opus uses a digital filter that moves the energy present after the cutoff frequency to the previous high frequencies;
How can this be an opinion? Either there's a filter or there isn't. And I'm not sure what you mean by moving energy. Filters don't move energy. They can amplify or attenuate energy at different frequencies, but never move it. Plus the Opus encoder doesn't apply an explicit low-pass filter. It just doesn't code the transforms coefficients that are too high (> 20 kHz).

Quote
it is a problem that depends on the digital low-pass filter chosen for Opus and not strictly on the Opus encoder. For this reason, it does not convince me when I compress songs that I have at a sampling frequency higher than necessary.
What does not convince you? Again, it's unclear what you're talking about, but whatever it is, it should be about facts and not being convinced of something.

Quote
However, to lower the sampling frequency, it is necessary to use a low pass filter, so apparently it might be better to use a different encoder. In fact, we always have files obtained by resampling the result of the manipulations performed at larger samplings and whenever the sampling or bit depth of a song is reduced, noise is inserted.
Yes, resampling involves a low-pass filter. And unless you're totally incompetent at implementing it, any artifact introduced by the resampling is orders of magnitudes smaller than the artifacts caused by the best lossy audio encoder. As for bit depth, you know you don't have to reduce it when resampling or encoding, right? Opus will happily take 24-bit float as input and preserve more than 24 bits of dynamic range.

 

Re: Classical/opera 192kHz/24bit to Opus

Reply #15
Don't dither.    Dither is supposed  to help when you reduce the bit depth but with lossy compression you're not storing the individual samples so there is no traditional bit depth.

...And, at normal-reasonable conditions at 16 bits or better you can't hear dither (or the effects of dither) anyway.

So if I want to convert 24 bit audio to a lossy codec that only supports 16 bit input I shouldn't bother dithering when converting it source audio from 24 bit to 16 bit?

 
SimplePortal 1.0.0 RC1 © 2008-2020