AUFTAKT is an automatic BPM (beats per minute) detection algorithm for recognizing the tempo of any audio input signal. Furthermore, the location of the beats is detected and represented along a timeline as a beat grid.
Tempo and beat location information are useful for a range of use cases, including:
- Playlist generation
- Auto-sync functionality for DJ software
- Automatic generation of high-level metadata information (e.g. ID-Tags)
- and much more
The most recent version of AUFTAKT, Version 4, relies on deep neural networks and state-of-the-art statistical models to take beat tracking to a new level of accuracy and reliability. It seamlessly integrates with our BARBEATQ automatic cue point detection SDK, or it can take beat information from a different source, such as a fixed beat grid.
The network has been trained with a diverse catalog of full musical compositions, ranging from Western Pop to Classical, and Electronic to more diverse music from around the world. The network was also trained on single instrument loops, including drums, guitars and pianos in order to expand the range of trackable music styles and performances.
AUFTAKT V4 improved detection rates by more than 30% across all genres, and even more than 40% on some styles, such as complex Jazz, compared to its previous versions. AUFTAKT is able to track tempo variations and changes, as well as time signature changes.
Features of AUFTAKT V4:
- Highly improved tempo detection and beat tracking ability
- Improved downbeat detection
- Ability to track abrupt tempo changes
- Ability to track time signature and time signature changes
- Capable of establishing multiple correction points for improved tracking
