I've been trying out three stem separation tools, designed to separate a complete mix into its different instruments.
Some comments about the some advantages and disadvantages of these three tools:
Spleeter: one of the first projects of this kind, initially developed by Deezer, who generously decided to release the code. It is the engine used by most of the paid online sites that offer this service, of which there are many.
Advantages: Free and open source, without any limitations. Offline, it can be used locally by computers without internet connection.
Disadvantages: Difficult to use, although it can be used without installing anything (the instructions are on Github), most people find command line programs too difficult to use (however for Windows users there is a version with a graphical environment and everything necessary already installed called Spleeter GUI).
GitHub - deezer/spleeter: Deezer source separation library including pretrained models. > Official website
SpleeterGUI | Maken It So > Version with graphical environment for Windows
Moises: originally an implementation of Spleeter designed to facilitate is use for musicians, they created a web service as an online interface service for Spleeter, and then they added some extra features (chord detection, automatic tempo, tempo and pitch modification, etc...), Modifying some parameters of the algorithm as well. Better known among the community of musicians than Spleeter, because they have invested money in promoting themselves mainly through youtube channels.
Advantages: ease of use, modern interface, musician-oriented, useful additional features, mobile versions.
Disadvantages: not free, cloud service, for online use only.
App Moises: la aplicación para músicos | Elimina voces y más
Demucs: another open source project which, to me, offers in its version 3 a slightly higher quality than Spleeter / Moises.
Advantages: free, open source and without limitations, quality of the results, offline processing, no internet access required to work.
Disadvantages: difficult to use, another command line program that requires the installation of Phyton interpreters, requires a powerful machine, otherwise it is slower than Spleeter.
GitHub - facebookresearch/demucs: Code for the paper Hybrid Spectrogram and Waveform Source Separation
Some comments about the some advantages and disadvantages of these three tools:
Spleeter: one of the first projects of this kind, initially developed by Deezer, who generously decided to release the code. It is the engine used by most of the paid online sites that offer this service, of which there are many.
Advantages: Free and open source, without any limitations. Offline, it can be used locally by computers without internet connection.
Disadvantages: Difficult to use, although it can be used without installing anything (the instructions are on Github), most people find command line programs too difficult to use (however for Windows users there is a version with a graphical environment and everything necessary already installed called Spleeter GUI).
GitHub - deezer/spleeter: Deezer source separation library including pretrained models. > Official website
SpleeterGUI | Maken It So > Version with graphical environment for Windows
Moises: originally an implementation of Spleeter designed to facilitate is use for musicians, they created a web service as an online interface service for Spleeter, and then they added some extra features (chord detection, automatic tempo, tempo and pitch modification, etc...), Modifying some parameters of the algorithm as well. Better known among the community of musicians than Spleeter, because they have invested money in promoting themselves mainly through youtube channels.
Advantages: ease of use, modern interface, musician-oriented, useful additional features, mobile versions.
Disadvantages: not free, cloud service, for online use only.
App Moises: la aplicación para músicos | Elimina voces y más
Demucs: another open source project which, to me, offers in its version 3 a slightly higher quality than Spleeter / Moises.
Advantages: free, open source and without limitations, quality of the results, offline processing, no internet access required to work.
Disadvantages: difficult to use, another command line program that requires the installation of Phyton interpreters, requires a powerful machine, otherwise it is slower than Spleeter.
GitHub - facebookresearch/demucs: Code for the paper Hybrid Spectrogram and Waveform Source Separation