New Opensource tool "Spleeter" for extracting stems

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.
So do you get the nice GUI version through purchasing the Acoustica Premium Edition; and a different command line version is the free one?

Acoustica has a GUI for Spleeter. They have a Spleeter add-on as a seperate download on their website. Acoustica has a 30 free trial (Premium). I’m trialing with that.

Spleeter in free but is command line only (if not using Acoustica). Spleeter doesn’t run in Windows it needs a Linux wrapper, this info on the Deezer (creators of Spleeter I think) Github webpage.

Efit: Acoustica has made a Windows addon version for use with Acoustica.
 
Acoustica has a GUI for Spleeter. They have a Spleeter add-on as a seperate download on their website. Acoustica has a 30 free trial (Premium). I’m trialing with that.

Spleeter in free but is command line only (if not using Acoustica). Spleeter doesn’t run in Windows it needs a Linux wrapper, this info on the Deezer (creators of Spleeter I think) Github webpage.

Efit: Acoustica has made a Windows addon version for use with Acoustica.
o_O... not sure I follow! So can I load Acoustica onto Windows 10; with Spleeter as an add on (is it like a VST?)
 
When you install the Acoustica, and try to to an stem extraction, the program will, with your permission, go out and get the Spleeter pieces it needs, then install them, and you won't miss a beat, or even have to know what it's getting or doing. It just does it. Boom.

One thing about this program. It's very fast.
 
There are posts about how to integrate spleeter in audacity, or by itself or as a vst plugin, but still nothing ready to use.
 
I tried this version, but couldn't get it to work. Maybe someone else will have better luck...

https://github.com/lazydevyo/SpleetGUI

AFAIK powershell is disabled by default on Windows 10, so that is strike 1. Strike 2 is that they are using pip, which if you used anaconda to install python (the "normal" way that you would to install spleeter), will likely just mess things up.

That said, I can try to get this going in a VM and post more detailed instructions.
 
When you install the Acoustica, and try to to an stem extraction, the program will, with your permission, go out and get the Spleeter pieces it needs, then install them, and you won't miss a beat, or even have to know what it's getting or doing. It just does it. Boom.

One thing about this program. It's very fast.

Acoustica has compiled Speeter in a custom Tensor.dll, and yes it's very simple and fast. Much faster than the "Spleeter" you get via python.

In a perfect world someone would figure out the interface for that Tensor.dll and then that could be used without the rest of/cost of Acoustica.
 
Acoustica has a GUI for Spleeter. They have a Spleeter add-on as a seperate download on their website. Acoustica has a 30 free trial (Premium). I’m trialing with that.

Spleeter in free but is command line only (if not using Acoustica). Spleeter doesn’t run in Windows it needs a Linux wrapper, this info on the Deezer (creators of Spleeter I think) Github webpage.

Efit: Acoustica has made a Windows addon version for use with Acoustica.

No. Spleeter runs on Windows, but is Python. It needs a Python distro. They don't develop on Windows, however, so Windows issues take longer to fix and don't get the love before a release.
 
AFAIK powershell is disabled by default on Windows 10, so that is strike 1. Strike 2 is that they are using pip, which if you used anaconda to install python (the "normal" way that you would to install spleeter), will likely just mess things up.

That said, I can try to get this going in a VM and post more detailed instructions.

Strike 3. It doesn't work on a fresh windows install, and you can see a lot of people are having the issue and it's not getting fixed:

https://github.com/lazydevyo/SpleetGUI/issues/1
I posted as well so maybe...

Also wasn't a good sign that the input dialog was expecting mp3 :0(

If you want to try this is what I did:

On a machine where you don't plan in install Adaconda or other Python distribution system;​
Run as administrator​
Click on Add Python to PATH at the bottom, then Install Now​
Run the exe as administrator​
After a bit you should see "Spleeter Installed" in Green at the bottom.​
1) pick your separation option (big Red/Grey squares)​
2) pick your output folder location (folder icon)​
3) Click on choose your file to select your input (it wants mp3 by default, not a spleeter thing, but this gui)​
4) Click split​
If there are errors you have to copy and paste the contents at the bottom (can't seem to enlarge or scroll to see the errors) into notepad or something.​
 
I wrote the attached batch file to split a folder full of flac files to stems - just copy the .bat file to the folder and double click. It creates both 2Stem and 5Stem sets using the 16kHz models, but could also do 4Stems with a minor tweak. I have noticed that the vocals on the 2stem model tend to be better than the 5Stem... Because Spleeter seems to have memory issues, this splits the source Flacs into small chunks for processing and then joins them back together afterward. You can delete the temp files once it is finished, but I have not set the batch file to do this automatically in case you have the need to use them. To install Anaconda/Spleeter in the first case, just follow the instructions here: deezer/spleeter. The batch script also uses SoX. You may need to adjust the paths at the start of the file to suit your own setup..
 

Attachments

  • Flac2Spleeter Stems (2+5).zip
    1.7 KB
I wrote the attached batch file to split a folder full of flac files to stems - just copy the .bat file to the folder and double click. It creates both 2Stem and 5Stem sets using the 16kHz models, but could also do 4Stems with a minor tweak. I have noticed that the vocals on the 2stem model tend to be better than the 5Stem... Because Spleeter seems to have memory issues, this splits the source Flacs into small chunks for processing and then joins them back together afterward. You can delete the temp files once it is finished, but I have not set the batch file to do this automatically in case you have the need to use them. To install Anaconda/Spleeter in the first case, just follow the instructions here: deezer/spleeter. The batch script also uses SoX. You may need to adjust the paths at the start of the file to suit your own setup..
I followed those instructions, in that I installed Anaconda, then downloaded the conda spleeter zip package but no clue from here. Help, anyone?
 
I wrote the attached batch file to split a folder full of flac files to stems - just copy the .bat file to the folder and double click. It creates both 2Stem and 5Stem sets using the 16kHz models, but could also do 4Stems with a minor tweak. I have noticed that the vocals on the 2stem model tend to be better than the 5Stem... Because Spleeter seems to have memory issues, this splits the source Flacs into small chunks for processing and then joins them back together afterward. You can delete the temp files once it is finished, but I have not set the batch file to do this automatically in case you have the need to use them. To install Anaconda/Spleeter in the first case, just follow the instructions here: deezer/spleeter. The batch script also uses SoX. You may need to adjust the paths at the start of the file to suit your own setup..

Cool, yeah as I said I had done similar for a friend but while it worked for me he couldn't get it going. Hopefully this will work for everyone.

Back to the SpleeterGui, I managed to get it going, but only after fixing the spleeter install and dependencies on the command line first. zcftr29's solution will be better, in terms of 16Khz models, however.

But if someone wants the gory details to get the GUI working I can post them.

The Gui also has C# source, so could probably be updated/fixed.
 
I followed those instructions, in that I installed Anaconda, then downloaded the conda spleeter zip package but no clue from here. Help, anyone?

You don't download anything. You should have an "Anaconda Prompt" icon on your desktop. double click that. Then in there you enter:

conda install -c conda-forge spleeter

for the cpu version or

conda install -c conda-forge spleeter-gpu

if you have a big Nvidia GPU (but I don't know if zcftr29's script supports that)
 
I started using this Spleeter GUI a few days ago and have been happy with it:
https://makenweb.com/spleeter2.php

Hmm, On my test VM (after unistalling all the the python stuff for the other GUI) I got:

Starting processing of all songs​
Processing C:\Users\IEUser\Desktop\01 - Stupid Girls.flac​
Traceback (most recent call last):​
File "C:\Users\IEUser\AppData\Roaming\SpleeterGUI\python\Lib\site-packages\numba\core\typeconv\typeconv.py", line 4, in <module>​
from numba.core.typeconv import _typeconv​
ImportError: DLL load failed: The specified module could not be found.​

I think a python package, numba, must have changed in the last day or so, because this GUI failed in (one of) the same way(s) that the other one did. Numba, therefore, needs to be downgraded to version numba==0.47.0.

However the python that is installed with this bundle doesn't have "pip", the command needed to change the numba version.

Researching...
 
OK if you get the numba error, I think what will work (I did some other stuff first but hopefully this is all you need):

1) type cmd in the search bar and click on "Command Prompt"
2) in the command prompt, type: cd %APPDATA%\SpleeterGUI\python
3) type: python -m pip install --upgrade numba==0.47.0

and then try the GUI again.
 
OK if you get the numba error, I think what will work (I did some other stuff first but hopefully this is all you need):

1) type cmd in the search bar and click on "Command Prompt"
2) in the command prompt, type: cd %APPDATA%\SpleeterGUI\python
3) type: python -m pip install --upgrade numba==0.47.0

and then try the GUI again.
That got it to work one time for me to do a two part split. Now it either returns nothing when it finishes or locks up/reboots the machine.
The "other" GUI once gave a message to use the Net configurator to allow something but othewise did not work.
No joy.
 
I have Python 3.8 installed on my Windows 10 PC, so I don't know if that's why I'm not having any issue using this GUI.
spleetergui.jpg
 
I have Python 3.8 installed on my Windows 10 PC, so I don't know if that's why I'm not having any issue using this GUI.
View attachment 50769

This one uses it's own python installed under %APPDATA%\SpleeterGUI\python and doesn't add any environment varibles, etc. so "should" be standalone. However I think it could get messy if you do have another python DOES add itself to PATH, etc.

I don't think the SpleeterGUI\python is a static part of the msi, or else we wouldn't be seeing these issues with the latest version of Numba, however. So, my theory is either that 1) You installed it BEFORE the broken for windows numba release (I think was on the 5th) or 2) Yeah your other Python is being used, because it's in your PATH environment variable. Guess "we" should look at the source code to lean more about how the GUI (and msi?) work, but I think the msi is separate and build instructions not shared. The author told me he doesn't know much about Python and his friend packaged it for him.

@boondocks; did you uninstall Anacoda before trying this one? Windows 10? How old is your computer?

For all these GUIs and Batch Scripts there has to be a working python and spleeter underneath, and unfortunately if you are not familiar with command line and some python commands it could be rough going if it doesn't work the first time/every time. Certainly for debug it would help to run spleeter on the command line and report any errors.

e.g.

1) Open a command window and cd %APPDATA%\SpleeterGUI\python
2) Type python -m spleeter separate -i "full path to a stereo wav file"
3) Capture text output (errors or success)
 
Back
Top