5.1 to 7.1.4 Source Separation Remixing

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

zeerround

Moderator
Staff member
Moderator
QQ Supporter
Joined
Apr 11, 2010
Messages
923
I was listening to Porcupine Tree's Fear Of A Dark Planet in 5.1 and it occurred to me to up/remix it to 7.1.4 using Source Separation.

Steven Wilson mixed this in a convenient way, in that there are no drums in the center channel, so I was able to do Stereo Source Separation on the front pair of channels, and the rear pair, to get 4 drum channels for the height speakers.

That would have been enough to give a 5.1.4 results but I wanted to go for 7.1.4 so tried another experiment that I think came out well.

I took the bass stems, resulting from the fronts AND rears stereo pairs, and routed them to the Side Surround channels in the 7.1.4 remix. At first this resulted in a mono bass sound, that felt as if it was coming from right overhead, and it wasn't as distinct from other instruments as I would like. So, I tried a producer's trick for widening a mono bass track and added a crossover at 180Hz (chosen by ear), routing the low end to the left side surround and high end to the right side surround. Now the bass sits in the "bed" layer of speakers, is distinct as a separate instrument source, and clearly in "stereo" as you get the growl from the left and the snap from the right.

Tools Used:

AudioMuxer to extract the 5.1 PCM track from the DVD and resample things when needed​
Plogue Bidule to make stereo pairs from the 5.1 (but other tools could have been used)​
Demucs for Source Separation​
Plogue Bidule to remix/remaster 9 stereo stems into the final 7.1.4 (front and back seperated sources and the original C and LFE stereo pair)​
A 7.1.4 capable version of my ebur128 measurement tool, to make sure the dynamics of the final 7.1.4 product were within 0.5 dB of the 5.1 source​
Here's what the remix/remaster PB looked like:

5.1 to 7.1.4 Source Separation Remix.png


And here is what the resulting 7.1.4 track (Let's Sleep Together) looks like in foobar2000:


Let's Sleep Together 7.1.4.png
From top to bottom the first 5 waveforms are the 5.1 "bed" layout, 6 and 7 are the side surrounds (stereo bass in this case) followed by the front height and rear height speakers (4 channel drums in this case).​
As I mentioned this track has only Steven's lead vocal in C. No drums or other instruments.​
For comparison, this is what you get with Penteo16 Pro 5.1 to 7.1.4:​
Let's Sleep Together Penteo 7.1.4.png
So nowhere near the "WOW" of separate instruments in different channels (plus it looks to be mixing signals from LF and RF into C).​
I'm happy to share the PB layouts if anyone is interested, although it does use some non-free VSTs for remastering, Namely SoftTube Drawmer S73 for analog warmth/character and to add back some "air" that gets lost with source separation.​
 
The next experiment was to tackle something that DOES have drums and other non vocal stuff in the 5.1 Center channel.

For this, I chose Soundgarden's Black Hole Sun, from the Superunknown Blu Ray. Heavy drums and more in the center channel.

So, I used demucs on three stereo pairs from the 5.1. Fronts, C&LFE and Rears, giving a total of 12 stereo stems. Bass, Drums, Vocal, and Other stems for each stereo pair.

The Fronts and Rears stems got the same treatment as in the Porcupine Tree above.

For the C&LFE stems the right channels of all stems (LFE in the 5.1) were fed to LFE in the 7.1.4.

The left channel of the bass stem (C in the 5.1) was fed to the crossover for delivery to the side surrounds in the 7.1.4.

The left channel of the Other and Vocal stems (C in the 5.1) was fed to the C channel in the 7.1.4.

The left channel of the Drums stem (C in the 5.1) was cut by 6dB then fed to both Height Front channels in the 7.1.4 (resulting in equal drum volume to the original).

Also a very positive result. The 5.1 had similar wave forms for fronts and rears, so the resulting waveform is not as dramatic as with Steven Wilson's 5.1 work, but still very pleasing as to filling the room with distinct sounds from different instruments coming from different channels.

1602965418523.png


Again, from top to bottom the first 5 waveforms are the 5.1 "bed" layout, 6 and 7 are the side surrounds (stereo bass in this case) followed by the front height and rear height speakers (4 channel drums in this case).
 
I was listening to Porcupine Tree's Fear Of A Dark Planet in 5.1 and it occurred to me to up/remix it to 7.1.4 using Source Separation.

Steven Wilson mixed this in a convenient way, in that there are no drums in the center channel, so I was able to do Stereo Source Separation on the front pair of channels, and the rear pair, to get 4 drum channels for the height speakers.

That would have been enough to give a 5.1.4 results but I wanted to go for 7.1.4 so tried another experiment that I think came out well.

I took the bass stems, resulting from the fronts AND rears stereo pairs, and routed them to the Side Surround channels in the 7.1.4 remix. At first this resulted in a mono bass sound, that felt as if it was coming from right overhead, and it wasn't as distinct from other instruments as I would like. So, I tried a producer's trick for widening a mono bass track and added a crossover at 180Hz (chosen by ear), routing the low end to the left side surround and high end to the right side surround. Now the bass sits in the "bed" layer of speakers, is distinct as a separate instrument source, and clearly in "stereo" as you get the growl from the left and the snap from the right.

Tools Used:

AudioMuxer to extract the 5.1 PCM track from the DVD and resample things when needed​
Plogue Bidule to make stereo pairs from the 5.1 (but other tools could have been used)​
Demucs for Source Separation​
Plogue Bidule to remix/remaster 9 stereo stems into the final 7.1.4 (front and back seperated sources and the original C and LFE stereo pair)​
A 7.1.4 capable version of my ebur128 measurement tool, to make sure the dynamics of the final 7.1.4 product were within 0.5 dB of the 5.1 source​


I have listened to these tracks and compared most of them to 5.1 to 12 channel upmixes I did using Penteo 16. Zeerround's version are definitely better. The isolation of the drums into the 4 ceiling speakers, opens up the fronts and creates more discreteness. Penteo doesnt put much info into the ceiling speakers.

Sound quality is also very good - much less artifacts than we typically find with stereo to 5.1 upmixes.

Compliments for the effort put into this!

Penteo 16 does have one major plus - it is very simple to use - drop all songs into Reaper batch tool and in one step generates the 12 mono wav files.

If Zeerround is able to automate most of these steps , preferably without Plogue, I'd definately be using it.
 
I was listening to Porcupine Tree's Fear Of A Dark Planet in 5.1 and it occurred to me to up/remix it to 7.1.4 using Source Separation.

Steven Wilson mixed this in a convenient way, in that there are no drums in the center channel, so I was able to do Stereo Source Separation on the front pair of channels, and the rear pair, to get 4 drum channels for the height speakers.

That would have been enough to give a 5.1.4 results but I wanted to go for 7.1.4 so tried another experiment that I think came out well.

I took the bass stems, resulting from the fronts AND rears stereo pairs, and routed them to the Side Surround channels in the 7.1.4 remix. At first this resulted in a mono bass sound, that felt as if it was coming from right overhead, and it wasn't as distinct from other instruments as I would like. So, I tried a producer's trick for widening a mono bass track and added a crossover at 180Hz (chosen by ear), routing the low end to the left side surround and high end to the right side surround. Now the bass sits in the "bed" layer of speakers, is distinct as a separate instrument source, and clearly in "stereo" as you get the growl from the left and the snap from the right.

Tools Used:

AudioMuxer to extract the 5.1 PCM track from the DVD and resample things when needed​
Plogue Bidule to make stereo pairs from the 5.1 (but other tools could have been used)​
Demucs for Source Separation​
Plogue Bidule to remix/remaster 9 stereo stems into the final 7.1.4 (front and back seperated sources and the original C and LFE stereo pair)​
A 7.1.4 capable version of my ebur128 measurement tool, to make sure the dynamics of the final 7.1.4 product were within 0.5 dB of the 5.1 source​
Here's what the remix/remaster PB looked like:

View attachment 58040

And here is what the resulting 7.1.4 track (Let's Sleep Together) looks like in foobar2000:


From top to bottom the first 5 waveforms are the 5.1 "bed" layout, 6 and 7 are the side surrounds (stereo bass in this case) followed by the front height and rear height speakers (4 channel drums in this case).​
As I mentioned this track has only Steven's lead vocal in C. No drums or other instruments.​
For comparison, this is what you get with Penteo16 Pro 5.1 to 7.1.4:​
So nowhere near the "WOW" of separate instruments in different channels (plus it looks to be mixing signals from LF and RF into C).​
I'm happy to share the PB layouts if anyone is interested, although it does use some non-free VSTs for remastering, Namely SoftTube Drawmer S73 for analog warmth/character and to add back some "air" that gets lost with source separation.​

Damn great job. I'm trying something like this. Have you done 5.1 source separation from Video source or Audio source? I wish to separate 5.1 Audio source but no forced separation like Mono in Audiomuxer but as it was/is in original mixed format. If Stereo as stereo or if C only Mono others stereo as it is.

How to go about it and What software you recommend? Even if paid software also if cost is nominal OK. My goal I want to upmix or Remix to my taste little bit with some plugins.

Kindly suggest
 
I was listening to Porcupine Tree's Fear Of A Dark Planet in 5.1 and it occurred to me to up/remix it to 7.1.4 using Source Separation.

Steven Wilson mixed this in a convenient way, in that there are no drums in the center channel, so I was able to do Stereo Source Separation on the front pair of channels, and the rear pair, to get 4 drum channels for the height speakers.

That would have been enough to give a 5.1.4 results but I wanted to go for 7.1.4 so tried another experiment that I think came out well.

I took the bass stems, resulting from the fronts AND rears stereo pairs, and routed them to the Side Surround channels in the 7.1.4 remix. At first this resulted in a mono bass sound, that felt as if it was coming from right overhead, and it wasn't as distinct from other instruments as I would like. So, I tried a producer's trick for widening a mono bass track and added a crossover at 180Hz (chosen by ear), routing the low end to the left side surround and high end to the right side surround. Now the bass sits in the "bed" layer of speakers, is distinct as a separate instrument source, and clearly in "stereo" as you get the growl from the left and the snap from the right.

Tools Used:

AudioMuxer to extract the 5.1 PCM track from the DVD and resample things when needed​
Plogue Bidule to make stereo pairs from the 5.1 (but other tools could have been used)​
Demucs for Source Separation​
Plogue Bidule to remix/remaster 9 stereo stems into the final 7.1.4 (front and back seperated sources and the original C and LFE stereo pair)​
A 7.1.4 capable version of my ebur128 measurement tool, to make sure the dynamics of the final 7.1.4 product were within 0.5 dB of the 5.1 source​
Here's what the remix/remaster PB looked like:

View attachment 58040

And here is what the resulting 7.1.4 track (Let's Sleep Together) looks like in foobar2000:


From top to bottom the first 5 waveforms are the 5.1 "bed" layout, 6 and 7 are the side surrounds (stereo bass in this case) followed by the front height and rear height speakers (4 channel drums in this case).​
As I mentioned this track has only Steven's lead vocal in C. No drums or other instruments.​
For comparison, this is what you get with Penteo16 Pro 5.1 to 7.1.4:​
So nowhere near the "WOW" of separate instruments in different channels (plus it looks to be mixing signals from LF and RF into C).​
I'm happy to share the PB layouts if anyone is interested, although it does use some non-free VSTs for remastering, Namely SoftTube Drawmer S73 for analog warmth/character and to add back some "air" that gets lost with source separation.​
Yes, I'd like to see the PB layouts if possible.

Thanks!
 
I have basically a custom layout per song. depending on how I want to route the stems, but they do all have similarities. e.g. crossovers for the bass stem that feeds the side surrounds etc.

The other thing is that I built a bunch of 12 channel groups of VSTs I have, to add back some analog "sparkle" and "heat". If you don't have those VSTs you'll get red boxes, but you could alt-click and drag them out of the layout.

Also some of my bidule SDK plugins, for monitoring and metering. These also use the Plogue "discrete" signal, so you need to use the Plogue named:

"C:\Program Files\Plogue\Bidule\PlogueBidule_x64-64.exe"

I think I may have packaged some of this up for another QQ user. Let me look...
 
Damn great job. I'm trying something like this. Have you done 5.1 source separation from Video source or Audio source? I wish to separate 5.1 Audio source but no forced separation like Mono in Audiomuxer but as it was/is in original mixed format. If Stereo as stereo or if C only Mono others stereo as it is.

How to go about it and What software you recommend? Even if paid software also if cost is nominal OK. My goal I want to upmix or Remix to my taste little bit with some plugins.

Kindly suggest

I'm not clear on exactly what you are asking.

There are free (but hard to use) and commercial music source separation tools out there. I prefer a combination of Spleeter/GUI and Demucs. Demucs for drums and Spleeter/GUI for everything else. Then you have to mix demucs drums and spleeter vocal, bass (and optionally piano) stems and subtract that from the original stereo to get a valid "other" stem.

Acoustica 7 Premium Edition and iZotope RX 7 music rebalance would be two examples of commercial software for source separation.

Yes I have done this for audio from video, extracted with audiomuxer. I don't recall putting Dolby Atmos or DTS:X audio back into a video however (unless you count the video output of audiomuxer), but do have the capability. Keep in mind that both of those encoders require an intel mac and are (well) over $1K. In the case of Dolby Atmos it pretty much also requires you to have Pro Tools Ultimate, also on intel mac and very expensive.

Without one of those encoders I don't think there is any way to get more than 8 channel audio back into a video, so really not doable for hobbyists.

5.1 or 7.1 Audio up/remix for Video is doable at home, however.
 
OK I've attached some example Plogue Bidule layouts and supporting files.

Please read the Readme file!!!
 

Attachments

  • Muisc Source Sepration up-remix example layouts.zip
    16.2 MB · Views: 61
I'm not clear on exactly what you are asking.

There are free (but hard to use) and commercial music source separation tools out there. I prefer a combination of Spleeter/GUI and Demucs. Demucs for drums and Spleeter/GUI for everything else. Then you have to mix demucs drums and spleeter vocal, bass (and optionally piano) stems and subtract that from the original stereo to get a valid "other" stem.

Acoustica 7 Premium Edition and iZotope RX 7 music rebalance would be two examples of commercial software for source separation.

Yes I have done this for audio from video, extracted with audiomuxer. I don't recall putting Dolby Atmos or DTS:X audio back into a video however (unless you count the video output of audiomuxer), but do have the capability. Keep in mind that both of those encoders require an intel mac and are (well) over $1K. In the case of Dolby Atmos it pretty much also requires you to have Pro Tools Ultimate, also on intel mac and very expensive.

Without one of those encoders I don't think there is any way to get more than 8 channel audio back into a video, so really not doable for hobbyists.

5.1 or 7.1 Audio up/remix for Video is doable at home, however.

Thank You so much. I think I'm going in the right direction. I just wanted to confirm this with an expert like you.As you have said I have Acoustica 7 Premium & those Plugins you have mentioned.

Able to mix to a certain extent to my taste. Still keep on improving .

Thank You
 
Back
Top