We introduce a framework for audio source separation using embeddings on a hyperbolic manifold.
We introduce a new source separation dataset, Divide and Remaster (DnR), for training and testing algorithms aiming at separating monaural audio signals into speech, music, and sound effects/background stems.
We present an interactive music source separation system allowing a user to specify the spatial location of the instruments of interest. We explore various conditioning mechanisms ranging from simple concatenation to positional encoding and AdaIN.
Thrilled to announce that our paper entitled "Deep Learning Based Source Separation Applied to Choir Ensembles" has been accepted to the 21st International Society for Music Information Retrieval Conference !
music information retrieval
Our paper entitled "HARP-Net: Hyper-Autoencoded Reconstruction Propagation for Scalable Neural Audio Coding" has been accepted to the IEEE Workshop on Applications of Signal Processing to Audio and Acoustics !
neural audio coding
In this second part, we explore network conditioning for source separation tasks, using the sources' fundamental frequency as control input.
Musical source separation for SATB choral recordings, performed in both the waveform and spectral domain.
Fun and creative music making app; with csJam, you can improvise, perform, sequence, and compose with a variety of synthetic and sampled instruments.
An iPad app exploring different spectral synthesis and processing techniques, such as convolution, morphing and cross synthesis from a practical perspective.
System designed in PureData, which assimilates and learns from musical performances using Markov models.
iPad app exploiting musical concepts such as question-answer, motifs, and transposition to generate music algorithmically in interaction with its user.