|
|
|
08-02-2009, 01:00 PM
|
#1
|
Human being with feelings
Join Date: Feb 2006
Location: Synthopia
Posts: 1,729
|
Cockos can have the ultimate resampling algorithm...for FREE!
The SINC resampler currently used in REAPER is a waste of CPU time - the only setting worth using is the 'extreme' preset that's *extremely* slow, producing rendered files that are of less-than-stellar quality.Using this preset for real-time playback is out of the question. People are begging for a better resampler
Now what if there's a resampling algorithm that's orders of magnitude faster than the current one and at the same time so good that it beats the best resamplers out there?
Imagine the 'perfect' resampler - something that rivals R8brain Pro / iZotope in quality and is so fast that it can be used for real-time playback as well
Sounds too good to be true? Nope,it gets even better:
Meet SoX (Sound Exchange) - a free, open source audio processing toolkit:
http://sox.sourceforge.net/SoX/Resampling
Its resampler is one of the best things ever created in the open source world
Tests at http://src.infinitewave.ca indicate that SoX is the leader in resampling quality,rivaling R8brain Pro and iZotope while being much faster,using orders of magnitude less CPU
Since SoX's code is licensed under LGPL,the resampler can be used in any commercial program,free of charge
EDIT: Some cat dude made a Foobar2000 plugin:
http://www.hydrogenaudio.org/forums/...howtopic=67373
Last edited by synth; 08-02-2009 at 01:26 PM.
|
|
|
08-02-2009, 01:06 PM
|
#2
|
Human being with feelings
Join Date: Jul 2006
Posts: 1,809
|
be silly not too I guess
Kind regards
Dave Rich
|
|
|
08-02-2009, 01:08 PM
|
#3
|
Human being with feelings
Join Date: Jul 2006
Posts: 12,480
|
Quote:
Originally Posted by synth
producing rendered files that are of average quality.
|
"Average"? That's bullshit. Quality is very good (although not "best"). Only render times are not cool.
|
|
|
08-02-2009, 01:17 PM
|
#4
|
Human being with feelings
Join Date: Feb 2006
Location: Synthopia
Posts: 1,729
|
Quote:
Originally Posted by Dstruct
"Average"? That's bullshit. Quality is very good (although not "best"). Only render times are not cool.
|
Fixed (I was talking about the SINC resampler in general)
Last edited by synth; 08-02-2009 at 01:20 PM.
|
|
|
08-02-2009, 03:48 PM
|
#5
|
Human being with feelings
Join Date: Feb 2008
Posts: 254
|
+1
yes please!!
|
|
|
08-02-2009, 04:11 PM
|
#6
|
Human being with feelings
Join Date: Sep 2006
Location: ocean mists
Posts: 860
|
Hey synth,
Good point. Some almost audible shadows cast back into the mix even with the best Reaper conversions. This could up the quality.
Would be nice if there was a Windows interface with Sox or a vegetable level incorporation into Reaper so that command line challenged dimwits like me could more easily navigate.
John
|
|
|
08-03-2009, 12:31 AM
|
#7
|
Human being with feelings
Join Date: Sep 2007
Posts: 642
|
I could use better resampler.
Maybe this is suitable addition to bug/feature tracker section?
cheers
|
|
|
08-03-2009, 03:35 AM
|
#8
|
Human being with feelings
Join Date: May 2009
Posts: 1,265
|
Quote:
Originally Posted by synth
Since SoX's code is licensed under LGPL,the resampler can be used in any commercial program,free of charge
|
Yes, but remember to meet the LGPL requirements the code needs to be implemented as a dynamically linked library and the source code of that must be released so it can be modified, and since currently the resampling of REAPER seems to be in the binary, Cockos would need to outfactor that part.
But all difficulties aside a big +1.
|
|
|
08-03-2009, 10:26 AM
|
#9
|
Human being with feelings
Join Date: Jan 2008
Location: CA
Posts: 991
|
Quote:
Originally Posted by synth
producing rendered files that are of less-than-stellar quality.
|
so you can really hear the difference, can you?
I'm gonna answer that and say NO. This is unperceived sound quality deterioration, that is around -50db above the threshold of human hearing and dips well below -100db in the audible range. You aren't gonna hear this under any practical circumstance.
That being said, if it is easy enough to implement, there is no reason not to get a higher quality algorithm, so folks can have their "peace of mind".
__________________
Damn it feels good to be a gangsta.
|
|
|
08-03-2009, 10:42 AM
|
#10
|
Human being with feelings
Join Date: Jun 2008
Posts: 4,923
|
StepOne - the problem is that the better algos in REAPER are too slow. the least good algos in REAPER produce degradation that is plenty audible....
|
|
|
08-03-2009, 10:48 AM
|
#11
|
Human being with feelings
Join Date: Feb 2007
Location: Oulu, Finland
Posts: 8,062
|
Quote:
Originally Posted by semiquaver
StepOne - the problem is that the better algos in REAPER are too slow. the least good algos in REAPER produce degradation that is plenty audible....
|
You have proof of that audibility? Blind testing done?
__________________
I am no longer part of the REAPER community. Please don't contact me with any REAPER-related issues.
|
|
|
08-03-2009, 11:24 AM
|
#12
|
Human being with feelings
Join Date: May 2009
Posts: 309
|
+1 for this, much needed an high quality algorithm that is also FAST
Not everyone can use offline rendering so a FAST algorithm that keeps quality is MUCH NEEDED.
|
|
|
08-04-2009, 04:10 AM
|
#13
|
Human being with feelings
Join Date: Jul 2007
Posts: 1,083
|
Quote:
Originally Posted by Xenakios
You have proof of that audibility? Blind testing done?
|
Come on! Always this real life stuff. Theory is so cool. We're talking about music in the end!
;-)
Shogger
|
|
|
08-04-2009, 04:18 AM
|
#14
|
Human being with feelings
Join Date: Feb 2007
Location: Oulu, Finland
Posts: 8,062
|
Quote:
Originally Posted by shogger
Come on! Always this real life stuff. Theory is so cool. We're talking about music in the end!
;-)
Shogger
|
Seriously, I am wondering why Cockos should apparently urgently waste effort on adding something of which the majority of users would fail in a blind test to even hear if there is any difference...
__________________
I am no longer part of the REAPER community. Please don't contact me with any REAPER-related issues.
|
|
|
08-04-2009, 04:42 AM
|
#15
|
Human being with feelings
Join Date: Mar 2008
Location: Sydney, Australia
Posts: 3,955
|
i like to think i have decent hearing, and i've had it set to 'good' quality (instead of the better quality ones) by mistake and not noticed it - at least, for a while. i did figure something was up with my session but it certainly didn't grab me, it was more like a slightly fuzzy loss of focus or something.
|
|
|
08-04-2009, 06:28 AM
|
#16
|
Administrator
Join Date: Jan 2005
Location: NYC
Posts: 15,721
|
Quote:
the least good algos in REAPER produce degradation that is plenty audible....
|
And these modes are provided for creative potential -- the medium quality modes sound great and are relatively quick, IMO.
We have no problem in adding support for LGPL algorithms (for example we use SoundTouch...). The main problem is that REAPER's SRC needs to be dynamic, i.e. supporting changing samplerates on the fly without big penalty.
|
|
|
08-04-2009, 06:30 AM
|
#17
|
Human being with feelings
Join Date: Jul 2006
Posts: 12,480
|
Quote:
Originally Posted by Justin
The main problem is that REAPER's SRC needs to be dynamic, i.e. supporting changing samplerates on the fly without big penalty.
|
And SoX doesn't support this (like r8brain)?
|
|
|
08-04-2009, 12:29 PM
|
#18
|
Human being with feelings
Join Date: Jul 2007
Posts: 1,083
|
I just resampled a 96kHz file to 44.1kHz. Once with "Good" and once with the "Better 192" algo. The phase cancelation test showed -80dB. I can't hear a difference.
Shogger
|
|
|
08-05-2009, 04:20 AM
|
#19
|
Human being with feelings
Join Date: Jun 2009
Location: Canada
Posts: 320
|
When does the SRC come into play, does it affect anything if all the files in the project are record or imported and then exported at the same sample rate.
|
|
|
08-05-2009, 04:44 AM
|
#20
|
Human being with feelings
Join Date: Mar 2008
Location: Sydney, Australia
Posts: 3,955
|
Quote:
Originally Posted by DaveG
When does the SRC come into play, does it affect anything if all the files in the project are record or imported and then exported at the same sample rate.
|
no - they never get used at all unless there's a sample rate mismatch
|
|
|
08-05-2009, 05:10 AM
|
#21
|
Human being with feelings
Join Date: Jun 2009
Location: Canada
Posts: 320
|
Quote:
Originally Posted by dub3000
no - they never get used at all unless there's a sample rate mismatch
|
Seems silly that they should process on demand using up CPU resources rather than just resampling to the project sample rate when imported. Am I missing something here.
|
|
|
08-05-2009, 05:21 AM
|
#22
|
Human being with feelings
Join Date: Mar 2008
Location: Sydney, Australia
Posts: 3,955
|
Quote:
Originally Posted by DaveG
Seems silly that they should process on demand using up CPU resources rather than just resampling to the project sample rate when imported. Am I missing something here.
|
because the sample rate is dynamic i.e. you can do smooth ramping of playrate and stuff i think. also saves generating lots of gigantic temporary files.
pro tools pre-processes audio on import and it's slow as hell. i think i prefer reaper's way. same with time stretching - reaper is way better in that regard (non-destructive time stretch yay). you can always manually glue items to lock them to the rate if you want to do it like that.
|
|
|
08-05-2009, 05:30 AM
|
#23
|
Human being with feelings
Join Date: Jun 2009
Location: South, UK
Posts: 14,214
|
dub have you so quickly forgot about sample rounding and null issues
I too would like to keep reaper as it is for sample rate stuff but need that sample rounding issue sorted !
|
|
|
08-05-2009, 05:41 AM
|
#24
|
Human being with feelings
Join Date: Jun 2009
Location: Canada
Posts: 320
|
Quote:
Originally Posted by dub3000
because the sample rate is dynamic i.e. you can do smooth ramping of playrate and stuff i think. also saves generating lots of gigantic temporary files.
pro tools pre-processes audio on import and it's slow as hell. i think i prefer reaper's way. same with time stretching - reaper is way better in that regard (non-destructive time stretch yay). you can always manually glue items to lock them to the rate if you want to do it like that.
|
Makes perfect sense now. Is this how ReaTune works as well.
Last edited by DaveG; 08-05-2009 at 05:44 AM.
|
|
|
08-08-2009, 03:47 AM
|
#25
|
Human being with feelings
Join Date: May 2009
Posts: 1,265
|
Quote:
Originally Posted by DaveG
Makes perfect sense now. Is this how ReaTune works as well.
|
No ReaTune uses pitch shifting algorithms such as Dirac, SoundTouch, or Elastique.
|
|
|
08-08-2009, 07:14 AM
|
#26
|
Human being with feelings
Join Date: Apr 2008
Posts: 510
|
Quote:
Originally Posted by Mich
Yes, but remember to meet the LGPL requirements the code needs to be implemented as a dynamically linked library and the source code of that must be released so it can be modified, and since currently the resampling of REAPER seems to be in the binary, Cockos would need to outfactor that part.
But all difficulties aside a big +1.
|
or improving what currently is available. but, by looking at the rmaa plot of the band transition, i'm not sure that doubling the n of points to 1024 for example will do much good. edit: can't really say i need such improvements
Quote:
Originally Posted by Dstruct
"Average"? That's bullshit. Quality is very good (although not "best"). Only render times are not cool.
|
yes, execution times of sinc filters, which are trying to approximate the perfect lp are really really slow. in fact its more of a modern cpu limitation, over an algorithm issue.
Quote:
Originally Posted by Xenakios
You have proof of that audibility? Blind testing done?
|
for a-b comparison of a single unmixed sample (a sine sweep) from two different src sources, differences would be clearly out of the audible range. however, it is arguable that in the case of a 16 tracks - 96khz mix, with 2 down sampling conversations in comparison, would produce audible differences. i cannot say for certain as i haven't done such test.
---
for free tools "r8b free" and "sox" should be considered superb and if you like to use them instead of reaper's src, do so.
however i'm not sure who declared "sox" to be the best (if we are talking the best of the best..etc)
window used: blackman, 1024bands, -220dbfs floor *note the lowered floor*, no img post-processing.
r8b free:
[img]http://img17.**************/img17/570/rb8freei.jpg[/img]
sox lin-phase vhq:
[img]http://img17.**************/img17/3075/soxlpq.jpg[/img]
my copy of cooledit from (y2000) :
[img]http://img17.**************/img17/4822/cedit.jpg[/img]
--
i've send this one to http://src.infinitewave.ca/ but they did not put it online, as it doesn't look very good or maybe because they are somehow biased :
traktor 2.5:
[img]http://img17.**************/img17/8866/tr261p.jpg[/img]
---
edit -> advice: avoid sample rate conversation as much as possible
lubomir
Last edited by liteon; 08-08-2009 at 08:10 AM.
|
|
|
08-08-2009, 09:00 AM
|
#27
|
Human being with feelings
Join Date: Jul 2007
Posts: 256
|
@liteon
- Those images are cool but... what the heck am I looking at?
__________________
I used to worry that I had cloth ears. Then I realized they were only painted on.
|
|
|
08-08-2009, 10:14 AM
|
#28
|
Human being with feelings
Join Date: Jun 2008
Posts: 4,923
|
Quote:
Originally Posted by Justin
And these modes are provided for creative potential --
|
it is a neat kind of distortion - fair enough.
but in certain situations I have to use the lowest methods for monitoring (when I am trying to record midi with low latency and there is a bunch of stuff being resampled, and in those cases if there was a more efficiant 'good' method that would be cool.
likewise when rendering stems and efficiency gains would be welcome of course...
but totally not a priority all things considered - the current setup is very workable.
and the 'dynamic-ness' of SRC in REAPER very cool - love the varispeed....
|
|
|
08-08-2009, 01:59 PM
|
#29
|
Human being with feelings
Join Date: Jan 2007
Location: in the middle of the icecube.
Posts: 7,403
|
I like all those pretty, blinky lights, but what do they sound like?
|
|
|
08-08-2009, 03:47 PM
|
#30
|
Human being with feelings
Join Date: Mar 2008
Location: Sydney, Australia
Posts: 3,955
|
They're all pretty close except for the last one which aliases like a bastard.
Aliasing sounds like weird harmonic distortion, where the distortion frequencies are sorta unrelated to the source frequency. Think something like running your mix through a ring modulator, but turned down a bit.
Last edited by dub3000; 08-08-2009 at 03:49 PM.
|
|
|
08-08-2009, 09:31 PM
|
#31
|
Human being with feelings
Join Date: Feb 2006
Location: Synthopia
Posts: 1,729
|
Quote:
however i'm not sure who declared "sox" to be the best free resampler
|
Do a phase test and you'll see where r8brain free fails.
But the biggest difference is in the SPEED of execution.
Last edited by synth; 08-08-2009 at 09:33 PM.
|
|
|
08-08-2009, 09:44 PM
|
#32
|
Human being with feelings
Join Date: Jun 2006
Posts: 22,567
|
nothing is EVER free
|
|
|
08-09-2009, 02:55 PM
|
#33
|
Human being with feelings
Join Date: May 2009
Posts: 1,265
|
Quote:
Originally Posted by Justin
The main problem is that REAPER's SRC needs to be dynamic, i.e. supporting changing samplerates on the fly without big penalty.
|
Would Cockos then consider licensing something like Secret Rabbit Code (aka libsamplerate)? Because it supports dynamic samplerate changes (at least it claims) and has better* (and/or same) quality than REAPER's current SRC while being faster and the commercial license is only 1000 AUD (~900 USD):
http://src.infinitewave.ca/
|
|
|
08-09-2009, 04:01 PM
|
#34
|
Human being with feelings
Join Date: Jan 2007
Location: in the middle of the icecube.
Posts: 7,403
|
if you ask me it would have to be significantly better sounding or significantly faster to even make an audible/tangible difference. I'd vote no to licensing a cross-grade quality.
greetings
.t
|
|
|
08-09-2009, 04:55 PM
|
#35
|
Human being with feelings
Join Date: Aug 2006
Location: Berlin
Posts: 11,817
|
Quote:
Originally Posted by Justin
And these modes are provided for creative potential -- the medium quality modes sound great and are relatively quick, IMO.
We have no problem in adding support for LGPL algorithms (for example we use SoundTouch...). The main problem is that REAPER's SRC needs to be dynamic, i.e. supporting changing sample rates on the fly without big penalty.
|
Perhaps the Optimizer (Schwa, the cycle hound) can improve efficiency of the SINC resamplers at some point.
Are render-caching strategies too messy to consider ? Ha! Would remind me all too well of Audiosuite processing, but without the shortcomings. The great thing about Reapers strategy is that you're ALWAYS working with the original audio. In PT we have to be quite careful not to reuse destructively time-stretched material in dialogue editing. The artefacts mount up quickly.
|
|
|
11-06-2009, 12:42 AM
|
#36
|
Human being with feelings
Join Date: Nov 2008
Posts: 24
|
If the highest audio quality is wanted every piece of gear and every process should produce artefacts way below known thresholds in order to avoid detection in high rez and "sensitive" set ups and combinations.
One must also realize that errors add up. You may not hear one source of degradation but you must consider the whole chain from mic to speakers and all that happens in between.
I export classical material out of Reaper and do SRC with SoX but I'd really prefer an algo in Reaper of similar high quality as SoX and iZotope to improve workflow.
I would like to see: higher quality SRC and several options of noise shaped dither in Reaper.
With my limited understanding of programming it seems like that would be relatively easy to implement and I can see no reason why.
I would be willing to pay 2-4 times the current price for these improvements.
Please Reaper team.. :-)
/Peter
|
|
|
11-06-2009, 01:36 AM
|
#37
|
Human being with feelings
Join Date: Jun 2006
Posts: 1,293
|
Couldn't someone code a SoX extension if it can't be added to the exe?
|
|
|
11-06-2009, 02:11 AM
|
#38
|
Human being with feelings
Join Date: Mar 2008
Location: Sydney, Australia
Posts: 3,955
|
the dither in Reaper is extremely good - you're not going to find much better than that.
if SoX can be invoked from the command line someone could throw together an extension for that pretty easily, i think.
|
|
|
11-06-2009, 02:12 AM
|
#39
|
Human being with feelings
Join Date: Mar 2008
Location: Sydney, Australia
Posts: 3,955
|
hmm, looks like it's cmd line friendly... let me see if i can throw something together really quickly.
|
|
|
11-06-2009, 02:32 AM
|
#40
|
Human being with feelings
Join Date: Nov 2009
Location: memory
Posts: 633
|
an src has a filter which needs to be set by ear.
the nyquist filter is a separate algorithm from the interpolation.
that is one reason that this subject is not so simple. the filter
should, imo always be linear phase, and that implies a f.i.r.
design, and that implies latency. i am not sure what kind of
filter is provided by s.o.x?
izotope has a f.i.r. that is configurable. however, this process will never happen
in real time (zero samples of latency) and still be sounding as good as it can.
this is not a matter of processor speed: when run offline, an src algoroithm can
process data at faster than realtime speed, because there is no need for playback...
but when a software-based f.i.r. filter is fed samples at normal playback
speed, it still requires a certain amount of samples latency to operate.
at least a couple of thousand in most cases.
so... there cannot be a true "real-time" src.
jeff dinces
Last edited by cerberus; 11-06-2009 at 02:42 AM.
|
|
|
Thread Tools |
|
Display Modes |
Linear Mode
|
Posting Rules
|
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
HTML code is Off
|
|
|
All times are GMT -7. The time now is 12:00 AM.
|