Getting started with upmixing

[ssully]

Is there a 'getting started' wiki/thread for a newbie? I see there's one at SurroundByUs, is that considered up to date?

 

[zeerround]

Surroundbyus has my stuff on it, and I check it daily, but you won't find commercial or other alternative tools there. It's also pretty much "crickets" these days so here is probably better.

If you want to try something free and drag and drop easy (on windows, stereo to 5.1 or 7.1) you can try SpecScript. SurroundByUs.com • View topic - SpecScript 1.8 <= Current version includes SpecWeb 2.4a6

If you are looking for something more than that (more than 7.1 or using Music source separation to up-remix), just ask.

 

[ssully]

Thanks. Just did the SpecScript install on a Win10 system and upmixed four tracks using 'Feeling Lucky'. Super easy. Output looks interesting in Audacity, will play it on an actual 5.1 system this weekend. Interesting that for the one stereo source that was a needledrop (done by me), it chose Mode 3 , whereas for the other three, from a CD of a different album, with good dynamic range, it chose Mode 5.

I see (and used) the 5.1-->4.1 postprocessing option too. But is there an option to NOT create an LFE at all? Either during or post-processing? I use bass management, not a fan of LFE in surround music mixes. though it's almost always there.

Am also interested in exploring Music source separation to upmix.

 

[zeerround]

Glad you made some quick easy progress!

Feeling Lucky looks at all the different modes and compares the channel to channel levels and picks the most even spread, so it's going to be different(track to track) based on how the stereo was mixed. I'm forgetting the gory details but there is also something about "weak center" from all the SpecScript methods selecting SpecWeb, vs. SpecScript.

Of course you are free to choose a method explicitly, and or do them all and audition for your favorite (for any given track).

Re: LFE, look in the "SpecScript 1.8 Release\bin\SpecScript.ini" file with a text editor (notepad++ or notepad) and set the lfe levels (each method) for say -110.0. That will give you 5.1/7.1 with an empty LFE. Don't forget double click RefreshValues.bat, in the same directory, in order for your changes to take affect.

For Music Source Separation there are LOTS of choses these days, but most are variations on models available from this free web site:

https://mvsep.com/en
That will get you started anyway, as far as Vocal, Drum, Bass, and "Other"

If you want to go for more separations, there are free models for piano, but for guitar (electric and acoustic) and more, you can try (paid but inexpensive):

https://www.lalal.ai/
If you have a BIG Nvidia GPU you can do stuff locally on your own computer, vs. using the web. Let me know if you are interested in that. but the results will be pretty much the same as MVSEP - Music & Voice Separation (plus a piano "stem").

Oh, there is also some models for splitting out echo and reverb. Lead/backing vocals is only in a commercial tool, but will be coming for free some time I'm sure.

if you use different tools/models for different stems, you'll need to make your own "other" stem by mixing all the stems together, inverting the phase, and then mix with the original stereo; (original stereo - (all stems) ) = other.

Once you have stereo stems for the different instruments/vocals, and other, you can decide which stems you want to upmix, and which you want to keep in stereo and just mix into 5.1 or 7.1.

Do any upmixing on the individual stems, then you can use audacity to remix everything.

Lastly there is thread here in this forum about a script method for up-remix, so you can look for that as well:

https://www.quadraphonicquad.com/fo...entrecutcl-stereo-to-5-1-script-v-0-2b.32788/
-Z

 

[madscot]

Sorry for jumping in but is there anything free or cheap as chips for Mac, Anyone? Love to upmix some of my favourite albums that probably wont ever see a 5.1 release

 

[zeerround]

Well you can run my stuff (SpecScript, etc.) in any of the ways to run windows on an intel mac. Virtualbox, parallels, etc. Not sure about M1 macs.

If all of that is too much complication we could start a thread on the upmixer in ffmpeg. Of course I prefer SpecScript but ffmpeg is there if nothing else. It needs a lot of parameters added. The defaults don't push things into the rears much, lower quality, etc.

 

[ssully]

My first experiment with SpecScript was promising. I converted the best mastering I know of Yes's Relayer (decoded HDCD from 1998) to 5.1 (mode 5, and also made a version post-processed to 4.1) and it came out quite nicely. Instantly more pleasing and 'right-sounding' than Steve Wilson's muffled thing. My only criticism so far is that it seems often pretty rear-biased, though that may just mean it's time for me to recalibrate my speakers, since I've been fussing manually with levels a lot recently.

What I'd like to do is compare a Dolby PLII (or DSU) upmix of a track, to SpecScript's. Has anyone done that?


(The Mode 3 conversion from an LP needledrop of a track a different album, on the other hand, had obvious artifacts (including during the silence before the track), and didn't sound nearly as good, as well as being even more rear-centered.)

 

[zeerround]

Re levels, those are all adjustable in the ini file.

The different methods will be different as to how the sound is distributed between the different channels.

I’m surprised to hear about artifacts as specscript has not had any reports of that, but please share the track details.

I can’t help you with matrix decode methods but I think you can get some via ffmpeg. There is also, no longer supported, the freesurround plugin for 32bit versions of foobar2000, which is said to be based on matrix decoding, but is also fft based like all my stuff.

 

[MidiMagic]

Discontinuitis!

Why bother with upmixing? Encode it inStereo-4, Dolby Surround, QS, or even SQ. I have done all of these.

 

[J. PUPSTER]

Discontinuitis!

Why bother with upmixing? Encode it inStereo-4, Dolby Surround, QS, or even SQ. I have done all of these.

Because it is a ton of fun!

Plus, IMO with todays tools one can get much better results than those Matrix systems.

I encourage folks to get into Up-mixing (I like to do what I call UDR or Up-mix/De-mix/Re-mix.)

Much of the music I love, that will never see a surround mix have really come alive again thank to Up-mixing; for me anyway.

It can be extremely time consuming to learn the craft of it and get good results, so it takes real commitment. I've been at it for several years now and learn new stuff every day; and feel I'm really only getting started.

If it is not for you that's perfectly fine.

 

[Sonik Wiz]

Discontinuitis!

Why bother with upmixing? Encode it inStereo-4, Dolby Surround, QS, or even SQ. I have done all of these.

Upmixing is not the same thing as encoding. Tell me how to take a stereo source & encode as QS.

 

[The56Kid]

Because it is a ton of fun!

Plus, IMO with todays tools one can get much better results than those Matrix systems.

I encourage folks to get into Up-mixing (I like to do what I call UDR or Up-mix/De-mix/Re-mix.)

Much of the music I love, that will never see a surround mix have really come alive again thank to Up-mixing; for me anyway.

It can be extremely time consuming to learn the craft of it and get good results, so it takes real commitment. I've been at it for several years now and learn new stuff every day; and feel I'm really only getting started.

If it is not for you that's perfectly fine.

I can attest to @J. PUPSTER’s upmixing artistry and, in my humble opinion, he’s able to achieve stunning results.

 

[ar surround]

Plus, IMO with todays tools one can get much better results than those Matrix systems.

I encourage folks to get into Up-mixing (I like to do what I call UDR or Up-mix/De-mix/Re-mix.)

If you are going to use Pups' UDR method, for the 'De-mix' process I highly recommend DeMix Pro 4.2. It is the only one of the stem separators that I am aware of that currently can separate lead vocals from backing vocals.

 

[Sonik Wiz]

Yes the @J. PUPSTER is a master re-mixer. I return to listen to his projects often. Upmixing can be fun but since I acquired the Surround Master I don't do much of it. The SM lacks the wrap around stereo synth mode that Sansui's had, so for something a bit special I alter the stereo original save as FLAC & play that back through the SM v3. I've mentioned this before but perhaps @ssully would consider this interesting:

https://www.quadraphonicquad.com/forums/threads/stereo-pre-synthesis.25299/
Alltho this was originally used as part of my own proprietary upmixing projects it can also be experimented with other upmixers or decoders.

 

[ssully]

I can’t help you with matrix decode methods but I think you can get some via ffmpeg. There is also, no longer supported, the freesurround plugin for 32bit versions of foobar2000, which is said to be based on matrix decoding, but is also fft based like all my stuff.

I'm just starting with the Freesurround DSP foobar plugin too. Reading the history of it Hydrogenaudio (starting in 2007) was a trip. While it seems to be an attempt to emulate Dolby Pro Logic IIx (released in 2003) in software, the author also seemed at great pains not to call it an upmixer. Instead there these convoluted use cases of downmixing a 5.1 mix, then re-upmixing it. Or starting from a DPL-encoded file and try to re-create the surround output without a Dolby decoder. And finally a strange mention of surround effects smuggled into stereo recordings. Weird.

Also, it appears there was at one time a Center Width parameter (as in DPLII) in FreeSurround but it seems to have gone away in the rebuild...unless I'm not really understanding the names of the adjustable parameters it does offer (which are several more than just DPLII's three, Center Width, Panorama, and Dimension). If anyone has a clue about this LMK.

 

[ssully]

(FWIW the reason I'm delving into DIY upmixers is because I miss DPL II Music mode)

 

[somethingcleveridunno]

If you are going to use Pups' UDR method, for the 'De-mix' process I highly recommend DeMix Pro 4.2. It is the only one of the stem separators that I am aware of that currently can separate lead vocals from backing vocals.

Second that if you're spending money on software. The vocal separation is insanely good.
I end up running the drum stem thru RipX as well, though, to separate the kick and the floor toms out from the other percussion.

 

[somethingcleveridunno]

I'm just starting with the Freesurround DSP foobar plugin too. Reading the history of it Hydrogenaudio (starting in 2007) was a trip. While it seems to be an attempt to emulate Dolby Pro Logic IIx (released in 2003) in software, the author also seemed at great pains not to call it an upmixer. Instead there these convoluted use cases of downmixing a 5.1 mix, then re-upmixing it. Or starting from a DPL-encoded file and try to re-create the surround output without a Dolby decoder. And finally a strange mention of surround effects smuggled into stereo recordings. Weird.

Also, it appears there was at one time a Center Width parameter (as in DPLII) in FreeSurround but it seems to have gone away in the rebuild...unless I'm not really understanding the names of the adjustable parameters it does offer (which are several more than just DPLII's three, Center Width, Panorama, and Dimension). If anyone has a clue about this LMK.

Despite the confusion, FreeSurround is pretty great as an upmixer! Some songs sound like they're deliberate discrete mixes when run through that, like Blue Öyster Cult's "Burning for You" or G&R's "Estranged"

Here's what the dev said about some of those settings:

• The sound field right after the decoding stage can be pictured as a 2-dimensional square with the listener at its center; this sound field can be further transformed spatially in several ways. The default setting is to leave the decoded field as it is, i.e., if a stereo track was downmixed from a discrete 5.1 track (e.g., from AC3 format) and is upmixed again with FreeSurround, it comes closest to the original source material when using the default settings. However, music that was not originally meant for surround playback can be spiced up quite a bit using these controls. The first option (Panorama) is to wrap the sound field around the listener in a circular manner. The slider allows to change the angle that the front soundstage takes up (i.e., the line between the front left and the front right corner of the sound field). Originally it is 90 degrees, but if it is set to, say, 270 degrees the front stage will be wrapped from behind the left ear along the front to behind the right ear of the listener. While the front stage is expanded, the side and rear sound field must correspondingly be compressed into the remaining space behind the listener. The second option (Dimension) allows to shift the sound field forward or backward without changing the shape. The third option (Depth) essentially scales the sound field to the back, i.e. the original square is resized along one axis (while the front stage stays where it is). These transforms are applied in the order in which they are listed here (the order matters). The last slider (Focus) is a bit experimental -- it allows to change the angular spread of the individual sound sources, either by making them more focal or less (i.e., more ambient). This is only really noticable with lots of speakers, but I would not overdo it.
• The second group of controls is how the sound field is mapped onto the speakers. The first setting here is the channel setup; this allows you to select how many speakers you have and where they are placed (front left and right, center, rear left and right, side speakers, etc.). From the point of view of the decoder there at 16 possible positions along the edges of the square sound field (see this picture for an example), namely the 4 corners, the 4 midpoints (front center, back center, left side, and right side), and two positions on each edge half-way between the corners and the midpoint (e.g., front left center, front right center). Most people have only a subset of all these channels (e.g., 7.1) in some common locations, but you can have some fairly arbitrary setups (***). But note that the placement of the speakers in the room is actually not in a rectangle but instead follows the corresponding home theater rules. If you have a subwoofer you might want to do the bass management right here, but it is usually a better idea to let the sound card or amplifier do it if they offer the option (e.g. for active sub-satellite systems). The last option in this box is the stereo separation; this allows to effectively stretch the sound field horizontally in the front and/or back areas.

I leave most of the settings alone but I crank up the localization all the way, turn on the bass redirect, set the crossover to match where my receiver's LFE crossover is set, and max out the stereo separation front and rear. Those sound best to my ears.

 

[Sonik Wiz]

Despite the confusion, FreeSurround is pretty great as an upmixer! Some songs sound like they're deliberate discrete mixes when run through that, like Blue Öyster Cult's "Burning for You" or G&R's "Estranged"

Here's what the dev said about some of those settings:



I leave most of the settings alone but I crank up the localization all the way, turn on the bass redirect, set the crossover to match where my receiver's LFE crossover is set, and max out the stereo separation front and rear. Those sound best to my ears.

Thanks for the info. I've Foobar 2000 installed but not the Freesuround plug in. I'm definitely interested. Can you tell me if this can be used to save & export a MCH file? Or is it only for real time playback?

 

[somethingcleveridunno]

Thanks for the info. I've Foobar 2000 installed but not the Freesuround plug in. I'm definitely interested. Can you tell me if this can be used to save & export a MCH file? Or is it only for real time playback?

You can use the Convert function (right-click on a track inside Foobar2000) to apply any DSP or combinations of DSPs, including FreeSurround, and save the resulting multichannel file as WAV or FLAC (or Apple Lossless if you install a free add-on codec pack). This can be a very effective way to convert matrix quad/surround files for playback on a surround system that lacks a built-in surround decoder.