Unmix Multiple Voices seems to crash (infinite loop, maybe) after a while

After chosing eleven voices to unmix, the process stopps at 46% and no further advancements for hours. SL has to be terminated via the taskmanager.

2 Likes

what kind of voices? singing or speaking?

Speaking. At most 2 people (male/female) talking at once. So it should be “fairly easy” for the algorithm.

NVIDIA GeForce RTX 4070
64 GB RAM
Intel I7-13700
Windows 11, up to date


Now, I tried it again with just one voice.
It stops at 44% and remains here.

1 Like

thanks for clarifying

are you able to separate 2 speaking voices? I mean, walk before running, no?

As I have said, I did similar…in my film, main contributor speaks with about five or six passers by…Unmix modules were unable to help me with that…I did it manually…and, the only speech that could have gone solo after my manual unmix was the main contrib…there was one guy who spoke a lot when no one else spoke…he got a supporting role :stuck_out_tongue_winking_eye:

if you watch @Unmixing video which he linked to the other day, he does point out that the DeNoise module which has a setting to separate voices is an experiment to try…I haven’t tried that yet…and my next job in SL is to remove some crew chatter…I might try that process

I just find that for my purposes, I lose far too much time with the modules as compared with manual editing

Are you sure the file isn´t some how corrupted? What happens, if you convert it to other format and then try it again?

1 Like

Please tell me, how the file is corrupt, when SL can do a lot of other stuff with it.
The file was created in audacity and saved as flac.
image


now, saved as wav
image

try to unmix 3 voices, stops working at 65%

And yes, when it comes to stability I do trust Audacity much more than SL at this moment right now.
I do use SL heavily and love it, but have to be careful of crashes anytime. That makes the use of SL a bit scary and unreliable (at this point).

1 Like

Then it has to be something in SL. Does mp3 version also stop? Worth to try, even though to sound quality is worse

Edit: What if you split the file in 2 separate files?

1 Like

Well, I would expect SL to work no matter what the external file-format is. Once the musicdata is inside the programm’s datastructures, It should not matter in any case. If it does, than this would be a rather serious issue, though.


When it comes to data structures, I wonder how the mapping between the actual music data and their representation on the screen is done programatically.

1 Like

Yes, I understand you, I´m just trying to think if there is a workaround to get your job done

1 Like

Hi…Just floating an idea & trying to help…
I’ve had a few problems with SL freezing/crashing when un-mixing difficult to make out voices & I solved it by cutting up the audio into smaller chunks, processing & then rejoining them - I’ve no clue if that is practical in your situation .
[ Be careful with De-Noise, it can can take out the ’ atmosphere ’ ]
Also.
There could be a problem with the Graphics Card - either not up-to-date or not powereful enough. If you are doing film editing, I presume you have good graphics but it might be a possibility. I saw some chap doing a demo on YT showing the processing difference with a high spec graphics card - no link , sorry.
Pardon me if you’ve tried these before.

1 Like

** Forther to my earlier reply. – i remember now - re- Graphics.
Try changing the AI Processing Device in Preferences / System from CPU to Graphics

1 Like

@paul_clegg IIRC, the GPU @Marc_von_Bredow is using is a bit long in the tooth

as is mine…I’m using a GeForce 1050 Ti and SL reports my GPU insufficient power if I try to switch to GPU processing

I think @Marc_von_Bredow has a point in that us users should think SL would at least complete the Unmix process at some point regardless of spec

that said, if user(s) don’t meet minimum spec of software, then said users should expect whatever software not to run :slight_smile:

Maybe it’s spend, spend spend & faff about one more time again ? 'It’s never ending.
Another method I tried [ with a dubious end result ] was to use the Magic Wand tool on the voice I wanted to isolate,& then trying to copy/paste to another track. This is rather clutching at straws though.
Last resort … ask Steinberg … :smile:

1 Like