Accepted at ISMIR: Wave-U-Net: A Multi-Scale Neural Network for End-to-End Audio Source Separation (with Daniel Stoller and Simon Dixon). arxiv.org/abs/1806.03185

Just published at the turn of the year: IEEE Signal Processing Magazine special issue in Music Signal Processing, featuring our article on automatic music transcription with Sebastian Ewert, Zhiyao Duan and Simon Dixon: ieeexplore.ieee.org/xpl/mostRecent…

It was great to have Daniel as intern at Spotify, or with the reviewer's words: "For the first time, lyrics alignment seems to be possible and can enable real-life applications" and "Obviously, the results are very impressive". Well done! Now on arxiv: arxiv.org/abs/1902.06797

Want to train a GAN but don't have much data?... Daniel Stoller, Sebastian Ewert, Simon Dixon "Training Generative Adversarial Networks from Incomplete Observations using Factorised Discriminators" arxiv.org/abs/1905.12660

We launched our newest feature SingAlong in Japan! Really amazing work by all teams involved! We integrated our latest published research on vocal separation and lyrics alignment, so that users can be a part of the song with highly accurate lyrics to support them.

Accepted at ICLR (with Daniel Stoller & Simon Dixon): If you ever wanted to train a CycleGAN (unsupervised) but you also had some paired examples and did not know how to maximize their usefulness (in a theoretically sound way), we have something for you: arxiv.org/abs/1905.12660

Soon at IJCAI: Ever wanted to use a strong non-causal architecture (U-Net) in autoregressive settings (wavenet, TCN)? Our Seq-U-Net not only accelerates training and inference but also enables multi-timescale modelling (D. Stoller, M. Tian, S. Dixon & I): sebewert.github.io/publications.h…

Soon at WCCI/IJCNN: Interested in optimizing source separation networks for mobile applications? We use the psychoacoustic model in any audio codec to increase the separation quality in resource-constrained networks (I. Ananthabhotla, J. Paradiso & I): sebewert.github.io/publications.h…

Spotify is hiring a Research Scientist to join our Audio Understanding team. For more infos, see here: spotifyjobs.com/search-jobs/#s…

Super proud of my team for this blog post (by Sebastian Ewert, Rachel Bittner, @nmontecc) covering some of our cutting-edge work at Spotify Research! 🎉 research.atspotify.com/making-sense-o…

I find yet again another submission that uses the test dataset to save the best model performance. Can we have a checkmark saying "I did not use the test set to decide which model to use" in our submission page please?

I'm super proud to announce the release of Pedalboard, a new Python library my team just open-sourced. Use VST and Audio Unit effects from Python in just a couple lines of code!

🎶 Spotify Research is looking for Ph.D. students for our Summer 2022 audio intelligence research internship (NYC or Europe) 🎧. Deadline November 15th! lifeatspotify.com/jobs/summer-in…

Accepted at IJCAI: Interested in Disentanglement Learning for sequences? For unsupervised local vs global disentanglement our new dynamic prior for global latents yields a simple and easy to train method that actually works (Yin-Jyun Luo, S. Ewert, S. Dixon):arxiv.org/abs/2205.05871

Meet BasicPitch.io — a 🎹🎸🎻🪕🎺🎷🪗 🎤-to-MIDI converter that uses about 70 billion fewer parameters than those fancy neural networks. engineering.atspotify.com/meet-basic-pit… #MachineLearning

We are very pleased to welcome Spotify Research as our Platinum and WiMIR patron sponsor for this year's International Society for Music Information Retrieval (ISMIR) Conference, to be held in Bengaluru, India from December 4-8, 2022, #ISMIR2022

Hey all! We're hiring a Machine Learning Engineer for our Audio Understanding team! Learn more here: lifeatspotify.com/jobs/machine-l…

We're super happy to receive the best industry paper award this year! Given that ICASSP had 6000+ submissions this years, we appreciate it a lot. IEEE ICASSP

Back underway in our 2023 #AIAwardsIrl finalist announcement as we reveal the #AI Person of the Year nominees. Congrats to all the finalists! Finalists are: Ciarán Gilligan-Lee, Spotify UK & Ireland Conrad Bernath, T-Pro Luca Longo, TU Dublin / OT Baile Átha Cliath Dr Maria Aretoulaki, GlobalLogic UK&I

If you missed it: Spotify is partnering with majors and indies to build AI products that put the artist first - Choice in participation - Fair compensation and new revenue - Focus on the Artist-fan connection We're hiring research scientists and engineers! newsroom.spotify.com/2025-10-16/art…