SoundsLike

Automatic recognition and description of musical properties

A piece of music can be described not only by its title, the artist or the album. It also has individual characteristics such as tempo, instrumentation, genre, or even the mood it creates. SoundsLike extracts these properties automatically from the music recording, offering versatile possibilities of using the data thereby gained.

Objective auto-tagging and efficient metadata generation

A manual indexing of large music libraries is both tedious and costly, which is why a lot of music datasets are incomplete or wrongly annotated. SoundsLike allows automatic annotation of musical data with objective metadata in short time. This annotation is based on intelligent algorithms, which are constantly developed further by means of large music databases. Automatic annotation via SoundsLike can effectively complement or even entirely replace the manual annotation.

Software solution for improved music search and recommendation

Based on automatic music annotation Fraunhofer IDMT has developed a software solution that allows the search for similar musical pieces. SoundsLike not only makes it easier to find certain music tracks in a database, specific search queries can also be implemented on the basis of individual criteria or automatically create music recommendations based on individual user preferences such as the current mood.

Automatic music recommendation and playlist creation

SoundsLike supports users in searching for pieces of music that fit for a certain purpose (advertising, video production, etc.). The fast creation of playlists is available by automatically recommending the most similar songs to a query music track. The similarity criteria can be individually weighted and adapted specifically to specific user requirements, allowing the  retrieval of less popular pieces of music in databases.

• Visualization of music databases
• Optimized archiving, made possible by detailed metadata for all tracks of a music database
• Intuitive navigation and search in large music databases of broadcasting stations, streaming services, or online music services, made possible by automatic indexing, search, and recommendation

• Genre (Classical, Country, Electronica, Jazz, Latin, Pop, Rap, Rock, Soul, Speech)
• Emotion (Anxious, Depressed, Exuberant, Content)
• Valence (High, Low)
• Arousal (High, Low)
• Perceived Tempo (Fast, Mid-Fast, Mid, Mid-Slow, Slow)
• Beats per Minute (e.g., 120 bpm)
• Key (C to H)
• Texture (Hard, Soft)
• Instrumental Density (Full, Sparse)
• Distortion (Clean, Gained, Overdrive, XXL)
• Dynamic (Continuous, Changing)
• Percussive (Nonpercussive, Percussive)
• Synthetic (Acoustic, Electroacoustic, Synthetic)

• Freely configurable compilation of metadata categories
• Tool can be adapted to meet user specific requirements (modular setup)
• SoundsLike music annotation and recommendation is available as a command line program and C ++ libraries for Windows, Linux and OS X , for being integrated with existing systems
• Test version available upon request