Innovation using #MusicBricks

11 seed prototypes were incubated using the #MusicBricks toolkit - ranging from data-enhanced product and hardware to SaaS.

Three were chosen to be incubated to prototype at #MTFCentral on the 18-20 September 2015 by judges who included some of the great music minds: Graham Massey (808 State), Stockhausen collaborator and composer Rolf Gehlhaar, Matt Black (Coldcut), producer and inventor Håkan Lidbo and British composer, recording artist and multi-instrumentalist Nitin Sawhney.

At the Sonar +D Music Hack Day in Barcelona, a special #MusicBricks workshop was held on the day before the MHD event. From the high number of projects that used #MusicBricks, four winners were chosen to develop their ideas further.

#MusicBricks tools made their premiere at #MTFScandi 28-30 May 2015. Four great seed ideas using #MusicBricks were chosen to be developed by our judges who included Warner Music’s Head of Technical and Creative (Digital) Josh Saunders, Coldcut’s Matt Black, vocal sculptor and recording artist Jason Singh, Deerlily Music CEO Paul Sonkamble and BBC Click presenter, musician and music hacker LJ Rich.

Read more about each of the incubated #MusicBricks projects below.

Gesture Sensors for Music Performance
The R-IoT sensor module embeds a 9 axis sensor with 3 accelerometers, 3 gyroscopes and 3 magnetometers, all 16 bit. It allows for getting 3D acceleration, 3-axis angular velocity and absolute orientation at a framerate of 200 Hz over WiFi. The core of the board is a Texas Instrument WiFi module with a 32 bit Cortex ARM processor that execute the program and deals with the Ethernet / WAN stack. It is compatible with TI’s Code Composer and with Energia, a port of the Arduino environment for TI processors. The sensor module is completed with a series of analysis MaxMSP modules that facilitates its use, based on the MuBu & Co Max library. This collection of analysis tools allows for: filtering and analyzing, computing scalar intensity from accelerometer or gyroscope, kick detection, Mdetection motion patterns such as “freefall”, spinning, shaking, slow motion. Further motion recognition tools are available in the MuBu & Co library.

Melody Extraction
This module includes a number of pitch tracking and melody transcription algorithms implemented in the Essentia library. Applications include visualization of predominant melody, pitch tracking, tuning rating, source separation.

Real-time Onset description
This module allows to detect onsets in real-time and provide a number of audio descriptors. It is part of essentiaRT~, a real-time subset of Essentia (MTG’s open-source C++ library for audio analysis and audio-based music information retrieval) implemented as an external for Pd and Max/MSP. As such, the current version does not yet include all of Essentia’s algorithms, but a number of features to slice and provide on-the-fly descriptors for classification of audio in real-time. A number of extractors analyse instantaneous features like the onset strength, the spectral centroid and the MFCC’s over a fixed-size window of 2048 points, after an onset is reported. Furthermore, essentiaRT~ is able to perform estimations on larger time-frames of user-defined lengths, and to report finer descriptions in terms of noisiness, f0, temporal centroid and loudness.

Sonarflow
A slick UI for browsing music by zooming into a colourful world of bubbles which represent genres, artists or moods which allows to discover new music online from various sources. It is available for iOS and Android, with APIs connecting to 7digital, last.fm, Youtube, Spotify, etc. Demo app in Google Play Store. See sonarflow.com and the source code on github.com/spectralmind.

Synaesthesia
A color detection tool triggering and controlling audio. Based on OpenCV it includes a colour detection engine and object tracking algorithm to control music and sound using plain colours or coloured objects. Using the camera it can trigger musical events. It includes an example app with a dedicated GUI. It is available for OSX and iOS. Find it on github: https://github.com/stromatolite/Synaesthesia

Freesound API
Freesound (www.freesound.org) is a state-of-the-art online collaborative audio database that contains over 200K Creative Commons licensed sound samples. All these sounds are annotated with user-provided free-form tags and textual descriptions that enable text-based retrieval. Content-based audio features are also extracted from sound samples to provide sound similarity search.  Users can browse, search, and retrieve information about the sounds, can find similar sounds to a given target (based on content analysis) and retrieve automatically extracted features from audio files; as well as perform advanced queries combining content analysis features and other metadata (tags, etc…).  The Freesound API will provide access to a RESTful API with API Clients in Python, Javascript, Objective-C.

Rhythm and Timbre Analysis
This is a library that processes audio data as input and analyzes the spectral rhythmic and timbral information in the audio to describe its acoustic content. It captures rhythmic and timbral features which can be stored or directly processed to compute acoustic similarity between two audio segments, find similar sounding songs (or song segments), create playlists of music of a certain style, detect the genre of a song, make music recommendations and much more. Depending on the needs, a range of audio features is available: Rhythm Patterns, Rhythm Histograms (i.e. a rough BPM peak histogram), Spectrum Descriptors and more. The library is available for Python, Matlab and Java.

MusicBricksTranscriber
(Melody & Bass Transcription + Beat & Key & Tempo Estimation)
The MusicBricksTranscriber provided by Fraunhofer IDMT is an executable that allows to transcribe the main melody and bass line of a given audio file. Also, the beat times, the key, and the average tempo are estimated. The results can be provided as MIDI, MusicXML, or plain XML files. In addition, a Python wrapper is included to further process the analysis results.

POF
POF = Pd + OpenFrameworks : openFrameworks externals for Pure Data, providing openGL mutlithreaded rendering and advanced multitouch events management. A recent addition to #MusicBricks by Antoine Rousseau, in collaboration with Matt Black of Ninja Tune, it makes making PD music apps much easier including cross platform. It also feeds the Mobile Orchestra through SyncJams. Find it on Github.

Musimap
Musimap’s algorithm applies fifty-five weighted variables to each music unit (e.g tracks, genres, labels) so as to model the world’s discography as a multi-layered system of cross-matched influences based on a musicological, lexicological and socio-psychological approach. The granular and proprietary database includes over 3B data points, 2B relations, and soon counting 50M songs. Its neural music network is the result of a unique combination of in-depth human curation and the latest AI technologies to engineer a multi-layered system.