Spleeter GUI - help
The good folks over at Deezer where nice enough to have released Spleeter for anyone to use freely, But python can be tricky to get working.
This Graphical User Interface project was created as a way to make Spleeter easier to use.
Spleeter is a Python application that uses an Artifical Inteligence called Tensorflow. It is quite resource hungry and
requires a modern computer to run.
Please do not email Deezer about problems with this project.
If you have any issues that are not resolved by the following help articles please create an issue here
https://github.com/boy1dr/SpleeterGui/issues and i'll see if i can help.
Intel Pentium & Celeron CPU's cannot run spleeter
If you are not running an intel i5/7/9 or Ryzen 5/7 or unsure if your CPU supports AVX, please use the
AVXcheck utility before trying to install or run spleeter.
Common problems
#1: ImportError: DLL load failed: A dynamic link library (DLL) initialization routine failed.
Failed to load the native TensorFlow runtime.
TLDR; TensorFlow is unable to run because your CPU does not support it.
There is a problem with the python package "TensorFlow", it is not able to run on your computer.
Your CPU model does not support AVX instruction sets.
Intel Pentium & Celeron CPU's cannot run spleeter.
With SpleeterGUI 2.5 and up there is a spleeter core update function in the menu bar at the top click "Help" then "Spleeter core upgrade".
This should make python's package manager complete the installation and correct any faults with the installation.
Some more information can be found here https://github.com/tensorflow/tensorflow/issues/31033
#2: The output folder contains separate part files but they are all the same.
SpleeterGUI is comes packaged with everything except the pre-trained model files, these are fetched when you first run each parts mode (2stem, 4stem, 5stem).
Once each of these have downloaded there are no further downloads required.
If all your output files are the same then this download process by Python has been blocked by anti-virus or firewall.
Ensure Python can download and try again.
#3: Can't load save_path when it is None
Similar to the above problem #2.
When python downloaded the pretrained model files something went wrong.
The fix is to delete them and run it again.
Go to C:\Users\<yourname>\AppData\Roaming\SpleeterGUI\pretrained_models
And delete the 2stem, 4stem, 5stem folders.
#4: VCRUNTIME.DLL was not found
Frequent questions
#1: Why is Version 2.6 a bigger download?
SpleeterGUI now comes with all files required to run.
Previous versions did not come with the pretrained model files because they added too much to the initial installer, but this proved to be problematic since Python needed permission to download and some anti-virus/anti-malware software blocked it (see common problems 2&3)
#2: Some output stems have a click in the beginning
There is a bug in the spleeter software that causes a click in the first frame of audio,
Deezer is aware of the problem and a fix should be coming soon.
Spleeter versions starting 1.5.1 up to current (at time of writing) 1.5.4 are known to have this.
When a fix is applied by Deezer you can upgrade your spleeter version by clicking "Help -> Spleeter core update".
If you would like to downgrade your spleeter version to 1.4.3, prior to the pop/click issue, download
this python folder
and replace the python folder here C:\Users\<YOURNAME>\AppData\Roaming\SpleeterGUI
#3: Is there a Apple Mac version?
Installing NVidia CUDA version
** ADVANCED USERS ONLY **
You are likely to run in to problems with this, be sure to google the errors you get to resolve problems for your hardware.
Settup up the NVidia environment
Settup up python/spleeter
Then run these commands from C:\Users\<yourname>\AppData\Roaming\SpleeterGUI\python
python -m pip uninstall spleeter
python -m pip install spleeter-gpu
At the time or writing, spleeter uses tensorflow-gpu 1.15.2
I will update this if spleeter updates the tensorflow package used.
CUDA versions for tensorflow
tensorflow-gpu |
CUDA |
cuDNN |
1.15.2 |
10.0 |
7.6.5 |
Remember, Google is your friend.
To undo the above install use these commands
python -m pip uninstall spleeter spleeter-gpu tensorflow
python -m pip install spleeter
Recomended computer specs
- Intel i5 (4th gen+)
- 8GB ram
- Windows 10 64bit
- Nvidia GTX1050 or better (if using CUDA method)
Tested hardware that works fine
- Intel i5 4460, 8GB ram, GTX1660ti
- Intel i7 7700, 16GB ram, GTX1060
- Ryzen 5 3600, 32GB ram, GTX1660ti
Not recomended
- Intel Celeron CPU
- Windows 32bit
- Windows 7
- Windows XP
- Computers purchased prior to ~2011
How to use
Download and run the installer. Then double click the SpleeterGUI.exe desktop shortcut to run it.
- Parts to separate (Vocal / Accompaniment / Bass etc)
- Recombine Choose which stems to mix back together (useful if you are learning drums and just want the drums removed from a bunch of songs)
- Full bandwidth: This enables full quality processing but results will vary depending on the song..
- Save to: Click the button and choose a place for the separated songs to be placed.
- Select 1 or more songs and drag them in to the application to begin processing.
For SpleeterGUI versions prior to 2.6...
The first time it processes a song there will be a delay while spleeter downloads some resource files.
If there are no files in the output folder or you see error messages then something has gone wrong. The most usual case is your computer has not met the minimum requirements to run spleeter.
If you are having troubles, Please check https://github.com/boy1dr/SpleeterGui/issues to see if the anser is there otherwise create a new issue and i'll check it out.