SPECWEB (Now 2.2)

[zeerround]

Hey if any of you feel like helping me get the word out on SpecWeb, there a couple of other forums you could post your experiances in:

http://forum.doom9.org/showthread.php?p=1753015#post1753015

and

https://hydrogenaud.io/index.php/topic,111046.new.html?PHPSESSID=foh4mrfqth3r4n04dl33hharf2#new

and

http://www.kvraudio.com/forum/viewtopic.php?f=6&t=455601&p=6361942#p6361942

Thanks,
Glenn

For some reason the links on doom9 and hydrogneaudio have changed. Try these:


http://forum.doom9.org/showthread.php?p ... ost1753015

https://hydrogenaud.io/index.php/topic,111046.0.html

I'm hoping if end users post their good results it will look less like spam and more like a ligit post that others would be interested in.

Thanks in advance.

 

[JerseyEd]

Since SpecWewb is a 32 bit program you use the 32 bit msi even on 64 bit OSs. Garry suggested that remove the 64 bit one to avoid confusion. The SpecWeb installer would have installed the 32 bit msi if you didn't already have it but since you had installed SpecWeb 1.0 you were already good to go for the Intel and MS redistributables (unless windows 10 deleted the intel one, as some have reported).

Re SBU discussion of SpecWeb 1.1, if you scroll down to the bottom of the release post, you will find a link to the discussion forum.

Thank you. Much appreciated.

 

[zeerround]

I have noticed however, when saving individual track settings to ini files from the web interface, that if you have Arctan+slice mode activated, it is not saved to the ini file. This is easy enough to fix in notepad prior to running the final conversion through the helper but I thought I'd mention it.

Yup, a bug. I have squashed it in the next release I'm working on now.

Thanks again for reporting it.

 

[zeerround]

I have noticed however, when saving individual track settings to ini files from the web interface, that if you have Arctan+slice mode activated, it is not saved to the ini file. This is easy enough to fix in notepad prior to running the final conversion through the helper but I thought I'd mention it.

Yup, a bug. I have squashed it in the next release I'm working on now.

Thanks again for reporting it.

 

[GOS]

So, I think I'm current. Downloaded version 1.1

I tried a single song from Grand Funk Closer To Home album and it sounded very good as an upmix.
So, I'm gonna go ahead and upmix the whole album.

 

[Neil Palfreyman]

Okay, here's some successes I have had in my first afternoon of using Specweb and the helper file:

Kate Bush "The Ninth Wave" - ie side two of "Hounds of Love" sounds incredible with default settings, though needs the rears lifting about 1db I was staggered by where some of the vocal effects were coming from on sections like "Waking the Witch" and "Jig of Life" - this had me sold hands down on specweb.

Steve Hillage "Green" This up-mixes quite heavily to the rears and also puts more percussion in the rears than I would ideally like, so needs more tweaking, but overall it sounds amazing! In particular Sea Nature, Ether Ships, UFO over Paris and Leylines to Glassdom sound super-trippy, with spacey synth arpeggios bouncing around the rears and Steve's guitar solos in the front.

Eloy "Planets" sounds absolutely immense with default settings.

Ozric Tentacles "erpland" was somewhat disappointing quite "thin" and inactive, but you can't win 'em all I guess.

 

[GOS]

OK, wow! I just let SpecWeb do an upmix of Grand Funk Railroad - Closer To Home. Seriously, it did a tremendous on this one. It really split things apart....very cool I gotta say.

 

[Neil Palfreyman]

Here's a couple more that seem to convert really well. I just did some test tracks so far (listed) but they're very active and hang together extremely well. May need some tweaking to take a little away from the rears in some cases, but very much worth the effort:

Hawkwind "Levitation" (Tested on "Levitation", "Motorway City", "Psychosis")
Roger Waters "Radio Kaos" (Tested on "Radio Waves", "Home", "Four Minutes" and "The Tide is Turning")

Sadly Hawkwind's "Chronicle of the Black Sword" didn't convert too well. Think I'll try some Greenslade next. "Pentateuch of the Cosmogony" in 5.1, anyone?

 

[Neil Palfreyman]

I'm pleased to report that "Pentateuch of the Cosmogony" by Dave Greenslade sounds amazing, too! I haven't had so much fun in ages. What next? Sigur Ros, Chicane, Dream Theater??? I'm like a kid in a sweet shop!

 

[zcftr29]

Hi all,

I'm having great fun and some spectacular results with SpecWeb. I was wondering if it would be good to set up a sticky of stuff that upmixes really well. It would be good if we could include catalogue numbers of the source and ini files with conversion settings used. I have so far had great success with the 2015 24/96 Qobuz versions of Muse albums (Showbiz, Origin, absolution, black holes) and the 24/96 versions of 2nd law and Drones. I've also done the first three Queen albums (2011 SACD versions) which have some really good rear separation. Is it possible to attache zip files with ini settings to posts on here?

 

[GOS]

Upmixing Miles Davis - Bitches Brew now...that should be interesting...

 

[zcftr29]

A question...

On the track 'Misfire' by Queen on the album 'Shear Heart Attack' [2011 SACD release] Freddie's vocals are mixed hard left and hard right, so they come out pretty discrete in the rear speakers - but it doesn't sound right. The rest of the album separates really well with lots of interesting stuff happening in the rears that you would expect there (harmony vocals, sweeping guitars, crashing cymbols etc). Also in Misfire, the centre speaker is dominated by the typical rear stuff. It seams the stereo field is somehow inverted. I tried a simple channel swap on the source file - left to right to see if this might be a mastering error but that just mirrored the sound along the central axis (as you might expect). Is there some way of pre-processing the input file so that hard left and right go to centre and visa versa? - maybe reversing the phase of one or both channels?

My ini settings are as follows:

;SpecWeb ini file format
;version 1.1
;
;ini file precedence order
;1) A file in the same directory as the input file, with the same name but .ini instead of .flac, or .wav etc.
;2) A file in the same directory as the input file, named SpecWeb.ini
;3) A file named SpecWeb.ini in the working directory of the icon properties
;4) A file named SpecWeb.ini in the same directory as the SpecWeb exe being run
;Only values from the first file found, in the above order, get used
;Any command line flags override the values from the ini file (that includes command line flags in shortcut properties)
;
;Comments are lines that start with ";"
;[section] defines a "section"
;parametername=value defines parameter values withen a section
;


[gain]

;0 zero pregain here = -12dB pregain in Plogue Spec. If bit depth is not 32f AND you clip, SpecWeb will run
;  again with a pregain calculated to avoid clipping.
;defaults all zero  (dB) 
; use -110 for LFE to turn off LFE (5.0 vs. 5.1) and speed up processing
pregain = 0.0
lf = 0.0
rf = 0.0
c = 0.0
lfe = 0.0
ls = 2.0
rs = 2.0


[method]

;slice, arctan, arctan+slice
mode = arctan+slice


[arctan]

;default 290
imagewidth = 290.0

;default 75
centerwidth = 80.0

;default 90
frontwidth = 90.0

;sum, across, diagonal default across
mode = across

;the amount of sound to include in each channel from adjacent speaker channels.
;0=none, default: 0.04 would be 4% adjacent 
adjspkr = 0.04

;The amount of sound from the blend source. 0=none, 0.05 would be 5% blend
;for ArcTan the blend source is the original stereo, for ArcTan+Slice, the blend source
;is the rears from Slice
;LS blend default 0.05 
lsblend = 0.30

;RS blend default 0.05 
rsblend = 0.30


[slice]

;A Humidity of 1 = 100% wet, or SLICE processed signal
;A Humidity of 0 = 0% wet or 100% dry signal from the previous stage
; (or original left and right in the case of stage one). 
;Stage 1 Humidity default 0.9
stage1humidity = 0.80

;Stage 2 Humidity default 0.95
stage2humidity = 0.95

;Wrap Rears adds a third stage of separation
;This has the effect of taking what was the extreme
;outside of the original stereo field and placing it
;in both rear speakers, creating a virtual center rear
;on, off default: off
wrap = off

;Stage 3 Humididty default 1.0
stage3humidity = 1.00


[fft]

;size is a a power of 2 >= 64 =< 32768
;default "auto" which will use the below sizes based on input file sample rate
; Recommended size for 48/44.1KHz input files = 8192. 96kHz input files = 16384.  176/192KHz input files = 32768.
size = 16384

;2, 4, 8, 16
;default 16 Higher number = higher quality and more cpu load
overlap = 16

;hanning, hamming, kaiser,
;default hamming
window = hamming


[zag]

;set to yes to use Z Automatic Gain, set to "no" for no Automatic Gain
;set to "normonly" to only normalize track so that loudest peak = your output setting
;(no channel to channel adjustments)
;default "yes"
usezag = yes

;match input song output level
;overrides output level setting if yes
;default=yes
matchlevel = yes

;peak output level will be (dB)
;default 0
outputlevel = 0.0

;center level, relative to fronts (dB)
;default 3
center2fronts = 3.0

;lfe level, relative to fronts (dB)
;default -9
lfe2fronts = -9.0

;rears level, relative to fronts (dB)
;default -3
rears2fronts = -3.0


[outputfile]

; mono, mch (6 mono file or a single 6 channel flac)
;default mch (6 channel flac)
channels = mch

;16, 24, 32, 32f, same (same=same as input, 32 and 32f are not compatible with flac
;output type will be changed to wav)
;default 24
bitdepth = 24


[browser]

;yes,no 
waitforbrowser = yes

;pathtobroswer is needed if your default browser is not Edge, Chrome, or firefox (e.g. if your default browser is ie)
;if your default browser is ie, install a supported browser and uncomment the bellow and check the path (or use "remote" and an iphone/iPad/Android device)
;pathtobroswer can also be "none" (without quotes), meaning no local browser on this computer, but wait for a remote browser to connect.
;pathtobrowser="C:\Program Files (x86)\Google\Chrome\Application\chrome.exe"
;portnumber for websocket server to use. default is 7681. if you have a firewall running on this computer is will need to allow SpecWeb on this TCP port
portnumber = 7681
pathtobrowser = default


[specwebversion]
current = 1.1
latest = 1.1

 

[zcftr29]

A question...

A possible solution...

I was thinking further about this on my may to work this morning and I have an idea.

If I set the sound field to 360 degrees, extreme left and right should both end up centre rear. If I also set the centre width to 0 degrees and adjust front width accordingly I should be able to get a Quad upmix with front speakers isolating the stuff I would ideally want in the rears and the rear speakers containing the vocals in the centre and the remainder spread between sL and sR. Swapping fL/fR for sL/sR should then get me close to something that sounds 'right'. I could then further process the new fL/fR with SpecWeb set to a 90 degree sound field and extract out the vocals to the centre channel...

I will try this out tonight, but any other suggestions would be most welcome.

 

[GOS]

A possible solution...

I was thinking further about this on my may to work this morning and I have an idea.

If I set the sound field to 360 degrees, extreme left and right should both end up centre rear. If I also set the centre width to 0 degrees and adjust front width accordingly I should be able to get a Quad upmix with front speakers isolating the stuff I would ideally want in the rears and the rear speakers containing the vocals in the centre and the remainder spread between sL and sR. Swapping fL/fR for sL/sR should then get me close to something that sounds 'right'. I could then further process the new fL/fR with SpecWeb set to a 90 degree sound field and extract out the vocals to the centre channel...

I will try this out tonight, but any other suggestions would be most welcome.

I'm clueless, but welcome to hear your findings. I have also found what seems to be excessive stuff in the rears that I might think goes in the front. Not often, but every once in a while.

 

[JanBakker]

Just did this one and it came out great, nice discrete stuff going on as you (maybe) can see in the waveform seekbar. This is getting addictive :yikes.

 

[skherbeck]

Just did this one and it came out great, nice discrete stuff going on as you (maybe) can see in the waveform seekbar. This is getting addictive :yikes.

I got similar results with "Diary of a Madman"... by the way, did you put those lyrics in manually or does your player "Grab them" from the internet somewhere?

 

[JanBakker]

Just did the rest of the album and if you have it, try it! It's really descrete at times and sounds way better than the stereo version ( still on my travel kit, dunno how it works out on my dear Onkyo )

 

[zeerround]

A question...

On the track 'Misfire' by Queen on the album 'Shear Heart Attack' [2011 SACD release] Freddie's vocals are mixed hard left and hard right, so they come out pretty discrete in the rear speakers - but it doesn't sound right. The rest of the album separates really well with lots of interesting stuff happening in the rears that you would expect there (harmony vocals, sweeping guitars, crashing cymbols etc). Also in Misfire, the centre speaker is dominated by the typical rear stuff. It seams the stereo field is somehow inverted. I tried a simple channel swap on the source file - left to right to see if this might be a mastering error but that just mirrored the sound along the central axis (as you might expect). Is there some way of pre-processing the input file so that hard left and right go to centre and visa versa? - maybe reversing the phase of one or both channels?

Well for the early Beatles stuff, with vocals in left channel only, people have swapped the LS channel with C. Your case sounds a little different, however, with vocal being in BOTH Left and Right but not the same signal in both. One thought (which is part of what you mentioned in your next post) would be to swap the fronts and rears.

By the way, you can ask SpecWeb to output 6 mono files, instead of one multichannel file, if that makes it easier to move things around. I know AudioMuxer can join mono's back into mch and I'm guessing sox and other tools can swap channels, etc.

It's not clear to me (without testing anyway) that setting the center width to zero will actually get you no center whatsoever. Spec and SpecWeb weren't designed for 4.0 or 4.1 and there is a lot of focus on the center channel in the code. The right way to create 4.0 or 4.1 would be to mix C equally into LF and RF. Also, unrelated to the current discussion, but specific to a true quad setup where you have speakers in the four corners of a square, Spec (the full version) has adjustments for speaker angles, so you can customize and tune your upmix for your specific speaker positions. SpecWeb (currently) has these angles hardcoded and assumes a 5.1 ITU setup. Again for Quad 4.0 or 4.1 the speaker angles should really be adjusted...

Anyway, back to your dilemma, try ArcTan Diagonal, as opposed to Sum or Across. Also try Slice with and without Wrap, as opposed to ArcTan. You may still need to swap channels but one of those may give better results on hard to convert otherwise tracks.

If you have 5.1 output on your computer you can play with all of this (except the channel swaps) live, using the SpecWeb play icon to play your upmix in realtime. You may need to edit the icon's properties if your default browser is IE (vs. real html5 browsers like Edge, Chrome, or Firefox) and the sound device to use if your 5.1 sound device is not device "1". Devices are listed with their numbers in the output you get when you drop a stereo file on the SpecWeb icon. When it says "enter options" you can scroll the window back up to see the sound devices and options listed.

No, I can't think of any pre-processing that would "invert" or turn the stereo "inside out". Easy enough to mixdown L and R into mono to get the vocals back to center, but then EVERYTHING will be in the center. Maybe some mixing of a little L into R and vise versa would make it less annoying? I guess you could try older extraction methods that involve subtracting channels too.

Rehase, Spec and SpecWeb work on the magnitudes of the magnitude/frequency representation of the spectral domain, so they are ignorant of phase info. All experiments with manipulating or using the frequency info (or translating it into phase and then using it) have pretty much come to naught, so no, changing the phase of the input won't do anything (other than phase interactions in your room due to things having funny phase).

There may be multi-tracks or stems for that track too?

 

[JanBakker]

I got similar results with "Diary of a Madman"... by the way, did you put those lyrics in manually or does your player "Grab them" from the internet somewhere?

Foobar grabs them from the internet and when available, highlights the frase being sung in real time, nice feature :banana:. I'll try Diary, thanks!

 

[Neil Palfreyman]

This afternoon I hooked my PC up to my AV amp so I can now play properly, tweaking the settings and saving .ini files as I go. Currently doing the title track of Jethro Tull's "Heavy Horses" and it sounds bloody amazing! ...going to try using the helper it up-sample it and use the ini file I created to process. This is great fun!