deMIX Pro is pretty damn good at vocal separations! Check this out

[Stephen W Tayler]

That's funny, because just before I read your post, I was flipping through the Azure d'or booklet looking at the credits. And there was your name! I won't ask you any more detailed questions as I understand your professional position. Also, as strictly a hobbiest/hacker, I probably will be clueless as to what you are saying anyway.

Well I can probably talk about Azure D'Or, it's just the projects I am working on now and all the ones that haven't yet been announced!! SWTx

 

[MixerRog]

Well, hopefully they get that crash event "deBugged Pro" soon. Would rather put my time into working the music rather than having to rig some work around.

I just ran a crash on a mono track, and then viewed the Win. 10 report in the Event Viewer, but didn't see anything that I could decipher (but of course I'm no Computer Tech guy.) I was thinking it could be some kind of .dll conflict with another program, but IDK.

As good as the program is, I can see lots of potential for improvement in the coming years

For those having this crash issue, I suspect it is likely an issue of your computer possibly running out of local memory. I suggest trying a song edited to 1 minute to see if it is ok. If it is ok, open task manager & run a separation of the full song & watch task manager to see if the memory is used up about the time of the crash.

Many AI programs require large amounts of memory or will crash. I recommend at least 16 GB.

Rog

 

[J. PUPSTER]

For those having this crash issue, I suspect it is likely an issue of your computer possibly running out of local memory. I suggest trying a song edited to 1 minute to see if it is ok. If it is ok, open task manager & run a separation of the full song & watch task manager to see if the memory is used up about the time of the crash.

Many AI programs require large amounts of memory or will crash. I recommend at least 16 GB.

Rog

Thanks for the suggestion Rog, I'll check that, but this is what I do most of my up-mix work from-
and I've never seen it crash a program since I got it, till dMP

Intel NUC 8i7BEK Mini Desktop, Intel Quad-Core i7-8559U Up to 4.5GHz, 16GB RAM, 512GB SSD, HDMI, Thunderbolt 3, Card Reader, Wi-Fi, Bluetooth, Windows 10 Pro

 

[MixerRog]

Thunderbolt 3

Do you have a GPU as the AI programs really like ones with 8 GB ram & a must have advanced Cuda to run locally fast? I don't have a GPU (too expensive but dropping due to now not needed for bitcoin mining) so I don't ever use local processing as way to slow but if you have that type of GPU, I have been told it should run pretty fast.

Rog

 

[J. PUPSTER]

Do you have a GPU as the AI programs really like ones with 8 GB ram & a must have advanced Cuda to run locally fast? I don't have a GPU (too expensive but dropping due to now not needed for bitcoin mining) so I don't ever use local processing as way to slow but if you have that type of GPU, I have been told it should run pretty fast.

Rog

Well we're probably going too far out of bounds for this thread, but I believe this little NUC only has an 'Intel Iris Plus Graphics 655' with 1GB ram.

I might want to some day switch to my newer computer that processes stuff about 3 times faster than the NUC with a high grade graphics card (but I don't recall what it is right now.)

I'll just stay confident that deMix Pro will fix the issue soon, and leave it at that.

 

[MixerRog]

Well we're probably going too far out of bounds for this thread, but I believe this little NUC only has an 'Intel Iris Plus Graphics 655' with 1GB ram.

I might want to some day switch to my newer computer that processes stuff about 3 times faster than the NUC with a high grade graphics card (but I don't recall what it is right now.)

I'll just stay confident that deMix Pro will fix the issue soon, and leave it at that.

If time is an issue for you like me at 72, it would be much to your advantage to use the cloud version of DeMix Pro 4 as it will be much faster I would think.

Rog

 

[ar surround]

Many AI programs require large amounts of memory or will crash. I recommend at least 16 GB.

I just tried a test on my Mac: 3.6 GHz Quad-Core Intel Core-i3; Radeon Pro 555X 2GB; 16 GB 2400 MHz DDR4 RAM; 500 GB SSD; macOS Ventura 13.1

I tried a 88/24 mono track using "local" separation...CRASH.

So we can't blame Windows 10.

 

[Stephen W Tayler]

I just tried a test on my Mac: 3.6 GHz Quad-Core Intel Core-i3; Radeon Pro 555X 2GB; 16 GB 2400 MHz DDR4 RAM; 500 GB SSD; macOS Ventura 13.1

I tried a 88/24 mono track using "local" separation...CRASH.

So we can't blame Windows 10.

Yes I already posted about this, and the company are aware, best to use "cloud" in the meantime.

 

[MixerRog]

I just tried a test on my Mac: 3.6 GHz Quad-Core Intel Core-i3; Radeon Pro 555X 2GB; 16 GB 2400 MHz DDR4 RAM; 500 GB SSD; macOS Ventura 13.1

I tried a 88/24 mono track using "local" separation...CRASH.

So we can't blame Windows 10.

As far as I know, the Radeon Pro 555X 2GB does not have the reguired Cuda & only 2 GB memory instead of 8 GB. I have the same Radeon Pro 555X 2GB with a quad core 3.2 ghz processor on Windows 10 & very slow with AI programs. In fact my on board video card was a little faster for some reason on Windows 10 but opposite on Windows 7! I even did a GPU software update with no change.

My friend has a 6 GB Cuda GPU & only takes seconds to process a song with most AI programs. He doesn't have DeMix Pro 4 yet to try it locally.

I do agree it could be a mono issue as I am a Beta tester & they have had some mono issues in the past. My computer is way to slow to test this issue on it.

Rog

 

[M-K]

I just tried a test on my Mac: 3.6 GHz Quad-Core Intel Core-i3; Radeon Pro 555X 2GB; 16 GB 2400 MHz DDR4 RAM; 500 GB SSD; macOS Ventura 13.1

I tried a 88/24 mono track using "local" separation...CRASH.

So we can't blame Windows 10.

Converting a mono to a (fake) stereo (or double mono) file is probably a workaround for this bug?

 

[Stephen W Tayler]

Converting a mono to a (fake) stereo (or double mono) file is probably a workaround for this bug?

Using "cloud" separation is the current workaround - yes creating a stereo file works, but that just makes more work if you want to convert it back to a mono file.

 

[M-K]

Using "cloud" separation is the current workaround - yes creating a stereo file works, but that just makes more work if you want to convert it back to a mono file.

I agree. Time is money. And it's an avoidable bug in my opinion.
And I can tell you there are many annoying bugs in professional audio software that I found and that will (probably) never be fixed. Pay and get the latest update (and new bugs as well)... I hate this.
Especially when it is not reproducible.

 

[J. PUPSTER]

Doing some demixing today and for some reason I'm getting the software hanging (like it just pauses with the circle work-action figure still turning at about the 80% done point); through the Local processing. Anyone else ever have this happen?

It still works with the cloud processing.

Maybe I need to re-boot and see if it fixes it. I'm on still on version 4.1.0 and says I'm up to date.

 

[MixerRog]

Doing some demixing today and for some reason I'm getting the software hanging (like it just pauses with the circle work-action figure still turning at about the 80% done point); through the Local processing. Anyone else ever have this happen?

It still works with the cloud processing.

Maybe I need to re-boot and see if it fixes it. I'm on still on version 4.1.0 and says I'm up to date.

Likely you are running out of memory on a long song. This program works best with a separate GPU with advanced CUDA
with 8 GB of GPU Ram. If you don't have anything close to that, I really suggest using their Cloud processing as usually much faster with the latest Stem processing updates also.

Rog

 

[ar surround]

Likely you are running out of memory on a long song. This program works best with a separate GPU with advanced CUDA
with 8 GB of GPU Ram. If you don't have anything close to that, I really suggest using their Cloud processing as usually much faster with the latest Stem processing updates also.

Rog

Is there some way of knowing when the Cloud separation software is a generation ahead of the Local algorithms in the current V 4.1?

 

[MixerRog]

Is there some way of knowing when the Cloud separation software is a generation ahead of the Local algorithms in the current V 4.1?

From what I know, the cloud updates are many times to speed cloud separations up but can be slight changes in separations also but anything major will be released as a official update.

Rog

 

[ar surround]

My newest adventure. Mike is in the front center. Mike's harmonies are front left and right. The group backing vocals are in the rears.

(I've still got to work on the "say you need me" to get the center voice moved to the left rear.)

 

[J. PUPSTER]

With news today of Jeff Beck, I just need to keep working; his taste and touch on guitar was unique and he remains one of my guitar Gods forever - R.I.P.

Started a new Up-mix (Penteo 4.1), de-mix (DeMIX Pro) and re-mix today sourced from the 2021 Japanese SHM-SACD of Camel's Moonmadness album.
song #5 Another Night.

DeMIX Pro did a great job of separating out stuff like the vocals, guitars, organ, drums; which I touched on all of that at one point or other, to highlight or move etc.

 

[ar surround]

Really clean separation of SM's lead and harmony vocals. It took very little work to do this one. Harmony vocals will be in the rears...or maybe front left and right. Decisions, decisions.

 

[J. PUPSTER]

Opened DeMix Pro to do some work today and it prompted me to download a new version 4.2.

Looks like it also retains the older 4.1 version Algorithm to choose from - (seems to run maybe 2x faster on 4.2 Local separation than 4.1!)