Splitter in it's basic functionality as a Deezer Spleeter web service with the standard 2 stem and 5 Stem models is 100% free and will remain free forever. No registration or email is required. We might add a few features here and there to make it fullfilling to you as the end-user. Ezstems is a website application that allows you to easily create audio stems from any audio file. The website converts audio files using Spleeter, which is a new artificial intelligence type of software.
Spleeter is the Deezer source separation library with pretrained models written in Python and using Tensorflow. It makes it easy to train music source separation models (assuming you have a dataset of isolated sources), and provides already trained state of the art models for performing various flavours of separation. The models available are. Recently, the research team at Deezer released a free and open source software as well as trained models to perform multi-source separation of music, with state-of-the-art accuracy. In this presentation we come back on our journey to open sourcing the Spleeter library, from doing the ground research, training the models, to releasing them.
Released on october 29th 2019, the Spleeter (https://github.com/deezer/spleeter) github repository received more than 5000 stars on its first week online and numerous positive feedbacks as well as press coverage. This talk will explain how we went from research code to this fairly easy to use open Python library, that integrates pre-trained models for inference and re-training.
While not a broadly known topic, the problem of source separation has interested a large community of music signal researchers for a couple of decades now. It starts from a simple observation: music recordings are usually a mix of several individual instrument tracks (lead vocal, drums, bass, piano etc..). The task of music source separation is: given a mix can we recover these separate tracks (sometimes called stems)? This has many potential applications: think remixes, upmixing, active listening, educational purposes, but also pre-processing for other tasks such as transcription.
The current state-of-the-art systems start to give convincing results on very wide catalogs of tracks, but the possibility of training such models remains largely bound by training data availability. In the case of copyrighted material like music, getting access to enough data is a pain point, and a source of inequality between research teams. Beside, an essential feature of good scientific research is that it must be reproducible by others. For these reasons and to even the playing field, we decided to not only release the code, but also our models pretrained on a carefully crafter in-house dataset.
Specific topics on which our presentation will dwell on are:- technical aspects of the models architecture and training- software design, and how to leverage tensorflow’s API in a user facing python library- how to package and version a code that leverages pre-trained models and that can be run on different architectures: CPU and GPU.- licensing and legal concerns- what we learned along the way- legacy
Stem Separation Audio
Deezer have just unexpectedly launched an amazing and powerful tool that they’re letting anyone use completely free of charge.
Deezer have announced ‘Spleeter’, their new free tool that can split stereo recordings into stem files. Spleeter claims to be able to split up the individual tracks from a sound recording, supposedly allowing you to access isolated channels for vocals, drums, guitars, etc.
To be able to split up a tracks elements after they’ve been merged together into one file is a much sought after but seemingly very difficult thing to do. Deezer compare Spleeter to the human brain, in how it is able to detect and isolate individual tracks from an overall mix.
They say: “Interestingly, our brain is very good at isolating instruments. Just focus on one of the instruments [in a track] and you will be able to hear it quite distinctively from the others.” So harnessing the idea of being able to distinguish track’s elements when we hear it, Deezer set about making that process a digital reality.
They continued: “That’s not really separation, you still hear all the other parts. In many cases, it may not be possible to exactly recover the individual tracks that have been mixed together. The challenge is thus to approximate them the best we can, that is to say as close as possible to the originals without creating too much distortion.”
Spleeter has many potential applications from remixing music using isolated tracks to research and education in music. It also pushes the area of Music Information Retrieval further forward, opening up the possibilities for other researchers to look at Deezer’s technology and see how they might be able to improve upon it even more.
Whilst it isn’t a perfect audio splitter it is impressive technology regardless and is capable of separating audio 100x faster than real-time. Spleeter is also MIT-licensed so it’s not only free to use it, but it’s free to use in any way you like however you will have to bear copyright in mind if you are using it on music you aren’t the rights-owner of.
Splitter Deezer Online
Deezer have released the technology to push research in the area of audio splitting. As they say themselves: “Spleeter is a neat tool, but in no way do we claim to have solved source separation. It’s our contribution to a vivid, ever-growing and open ecosystem and hopefully something others will build upon too.
“Finally, it’s worth pointing out that music mixing is a fine art and that mastering sound engineers are artists in their own rights. Obviously we do not intend to harm their work in any manner or affect anyone’s credit. When you use Spleeter, please do so responsibly.”