SpecWeb 2.3 Beta 1

[zeerround]

I don't think I've posted a beta version here before but let's see how it goes.

Here is SpecWeb 2.3 Beta 1

Major Changes from 2.2 Public Release:

Fix for faint "tick" sound (like dust on vinyl) during up/down sample

Improved transient smearing artifact reduction ("swishy" drums/cymbals) via new -y flag

From the SpecWeb.ini file:

;Starting with SpecWeb 2.3, arctan processes any "transient" sounds via the slice (-M0) method, to eliminate artifacts
;such as "swishy" drums/cymbals, etc.
;The threshold to determine what is transient is set with the below transientthreshold value.
;This is the slope of the sound change in volume at a given frequency and instant in time, so smaller values
;will cause more of the sound to be treated as transient (fewer artifacts) and a larger value will cause less of the sound to be treated as transient (more separation). The default value is 0.04

A goal of this Beta is to gather end user feedback on the default setting for Transient Threshold.

Beta?

Beta version means:

Changes may not be fully documented
May run a little slower (while I work on speed vs. stability)
No other bug fixes from the previous version (yet)

https://drive.google.com/file/d/1eVBJ64 ... sp=sharing

For older computers (pre AVX instructions) Updated 1-15-2021:

https://drive.google.com/file/d/1bqTd-ZcNVJtKSqkGrJBggo8S8r73RlfZ/view?usp=sharing
So, in return for early access to the new features, I'd like feedback from you on how this sounds, relative to 2.2, and what value for -y should be the default (if not 0.04 which is the default in this version).

FYI you can multiple versions of SpecWeb on your computer at the same time.

 

[zeerround]

Crickets...?

Disappointing.

Oh well.

FYI the No-AVX version isn't really No-AVX so its currently broken for CPUs from before 2011. I'll work on it but this stuff get's harder and harder as time goes on. Intel and Microsoft updates their tools...

 

[Mcallister]

Just saw this. I’m out of town will check it out as soon as I return.

 

[zeerround]

Post at top of thread updated with new link to fixed NO-AVX version, for older computers.

 

[BJMarsh]

I would love to try this but I am running a MacBook Pro. Is there a macOS version?

 

[zeerround]

Unfortunately, there is currently a ton of windows specific code in SpecWeb. It should run OK on an intel mac with Parallels or other virtual host software, bootcamp for booting into windows 10, etc.

I have no idea about M1 macs.

Oh, and there is also the "Spec" in Plogue Bidule approach. Where there is a Mac version (Requires a Plogue Bidule license). It hasn't been updated in a while but it is there on surroundbyus.com.

Running in Parallels or VirtualBox, etc. is probably your best bet. I successfully purchased Windows 10 licenses for $8 US in those cases.

 

[BJMarsh]

Ok, I have Parallels. I will do that. Thanks!

 

[EMR]

I would love to try this but I am running a MacBook Pro. Is there a macOS version?

I have an old Mac Pro 5,1 running the latest release of Big Sur (yes, I know in theory that shouldn't be possible). I can confirm I've been using SpecWeb 2.2 just fine (the version for old processors) on a Windows 10 VMware Fusion 11.x and 12 virtual machine in Catalina and on a Windows 10 Parallels Desktop Business Edition 16.1.2 virtual machine in Big Sur.

 

[EMR]

@zeerround
Is trashing version 2.2 and installing 2.3 beta in its place recommended?

 

[zeerround]

It shouldn't matter. You can have multiple versions installed at once. The only overlap would be the shortcuts on the desktop.

For instance if you wanted to keep an old version you could just rename the shortcuts, such as adding a version number, and then install the new version.

 

[beerking]

OOPS.....How did I miss this thread!!?

Will get on to it and report back.

 

[BJMarsh]

Got it going on my MacBook Pro with W10. Had to install 2.2 first since 2.3 was missing some DLLs. All good after that. Pretty impressed with how discrete it is. I do have artifacts when listening to the rears with headphones but when folded to stereo they were not audible. I’ll check it out on my 5.1 system. Also note I took all the defaults. Maybe I can tune some of the artifacts out.

My preference is to playback in real time. Anybody doing that?

 

[BJMarsh]

Got it going on my MacBook Pro with W10. Had to install 2.2 first since 2.3 was missing some DLLs. All good after that. Pretty impressed with how discrete it is. I do have artifacts when listening to the rears with headphones but when folded to stereo they were not audible. I’ll check it out on my 5.1 system. Also note I took all the defaults. Maybe I can tune some of the artifacts out.

My preference is to playback in real time. Anybody doing that?

 

[zeerround]

What Dlls are missing? It can't really be DLLs in the folder but rather something installed, e.g. MS C++ runtime or Intel C++ runtime.

Did you run the installer in 2.3 (before 2.2)?

I thought I tested from scratch in a VM, but thanks for finding a problem.

Re: listening live, at some point, it will start to sound bad. There's a thread on that here somewhere and I had posted that is was fixed, but I later had to back out the changes and haven't gone back to it. At the moment I don't remember the timing but it is several hours (depending on the sample rate). If you run into that just restart SpecWeb.

Re Artifacts, I remember posting about that as well. but yes adjusting the center and front widths (which also adjust the back widths) so that sounds are firmly in a given channel vs. spread between them would be one thing to try. Then you have various blend options, to either blend in some of the original stereo, or sound from an adjacent channel, or even from a different separation mode. E.g. ArcTan with Slice Blended Rears.

Check out the "How to win with ArcTan" section of the guide, mostly written by the infamous/famous DKA. Originally written for the Plogue Biddle "Spec", predating SpecWeb, it still has some application, although it also predates the now default "Zone2" mode.

If you tell me more about what "artifacts" you are hearing I may be able to be more helpful.

 

[BJMarsh]

The DLLs were not part of your distribution for either 2.2.or 2.3. - I checked the DLLs on both distributions. I suspect that the correct DLLs were loaded by the Intel or MS installers. Anyway, from memory - libmmd.dll, libio5md.dll and a few others.

The artifacts I hear are like when overdoing noise reduction with a small FFT. Kind of swishing on the edges.

 

[EMR]

All right, I’ve done a little experiment with SpecWeb 2.3beta1 and am going to report the subjective results. I must point out that the type of music I listen to is probably not the best to determine how good the improvement in the current beta is. I’ve tried SpecWeb on two very different types of stereo recording. In no case did I notice any artifacts.

One is the aria “I Dreamt I Dwelt In Marble Halls” (from M.W. Balfe’s opera The Bohemian Girl), as interpreted by mezzosoprano Elīna Garanča. I must point out that, for some reason I can’t really understand, Ms Garanča’s recordings are not really amenable to normal surround treatment, as SpecWeb does little more than add some ambience to the recording. Not so with recordings of other opera singers. For instance, recordings of Plácido Domingo and Franco Corelli sound truly 3D after undergoing the SpecWeb treatment. Others, like Alfredo Kraus, Luciano Pavarotti and Elīna Garanča, simply don’t.

So, in my subjective opinion, I didn’t perceive any significant differences between 2.2 and 2.3beta1 with that aria. I asked my wife to listen to both versions and she thinks that version 2.3beta1 might be marginally better in the best of cases. The audible differences (if any; I know the waveforms are different) are minimal.

The other song was the Engelbert Humperdinck’s recording of “If Tomorrow Never Comes”. Contrary to what happens with Ms Garanča, Humperdinck’s recordings sound great in surround, and this case is no exception. It sounds fantastic after the treatment with SpecWeb 2.2; version 2.3beta1 is also great. I can’t actually determine which version is “better”.

If you have any suggestions for me to try, I’ll gladly try other jazz and classical recordings, but perhaps my ears won’t be able to tell the difference.

 

[zeerround]

The differences between 2.2 and 2.3 are:

No more clicks (like dust on a record). Almost none of us noticed/reported that, however, even though it actually happens regularly. "We" only heard it in certain quiet passages.

A reduction of swishy artifacts on noisy transient sounds like snare drums/high-hats etc.

So that could explain why you don't hear a difference. The core functionality is the same.

I can point you to some tracks that clearly demonstrate the issues with 2.2, if you like (but they won't be Jazz or Opera).

Re: mixes that don't separate well, or only have ambiance in the rears (with the default settings), SpecWeb works by "stretching" the original stereo mix into 5 channels, and you have control over the "stretch" with the image and channel width controls. It also puts any sound that is mono (the same in both Left and Right, or another way to say it would be "panned center" in the mix, into the center channel, subject to the center width control.

So, a stereo mix that is mostly toward the center, with only ambient sound panned hard left and right, is going to have only ambient sounds in the rears, UNLESS, you change the width settings to stretch any mid centerish sounds into the rears. You don't actually have a rear width control, the rears get whatever is left from the total image width, after you set the center and front widths.

You could do this:

Adjust the center width to a minimum value to where the sound in that channel sounds full/complete by itself (you can solo the center channel), then do the same with the fronts.

Smaller values for center and front widths will put more sound to the rears.

Going back to what I wanted to get out of the beta is IF you hear swishy artifacts, at what setting of the "Transient Threshold" do they go away, but you still have good "ArcTan" separation on the non transient sound and/or the percentage of the track processed as transient (printed out at the bottom after the track finishes) is not overly high (say more than 20%?).

I hope that's clear. I'm surprised I didn't state that more clearly in this thread or in the Readme.

 

[EMR]

Fascinating. I knew parameters existed that allowed one to fine-tune the result of SpecWeb processing, but I was too lazy to look into that. I've read your response three times already and, apparently, I've succeeded in setting center width as 0 degrees by entering "-c0". However, I'm still confused with the width of the fronts. If I'm not mistaken, the default width is 36 degrees. So, I've tried setting a smaller angle by entering "-f18" and "-f5", but, in both cases, the result appears to be the same as not entering a "-f" parameter. I'm sure I'm doing something wrong, but I have no idea what that might be.

 

[zeerround]

No that should work. Have you considered using the "play" option, so you can monitor, make changes to parameters and solo/mute channels in real time?

 

[EMR]

Hmmm. I see no difference between -f5, -f18 and not entering anything at all. I'm at a loss. As for using "play", I don't think that would be useful in my case, as I don't have a laptop computer, and neither do I have a 5.1 system anywhere near my desktop Mac Pro.