Melody Extraction for Music Games

With the automated Melody Extraction of the Fraunhofer IDMT, you will receive innovative software that quickly and reliably detects the melody even under difficult acoustic conditions – no matter if it is a sung melody or a recorded audio track. This tool is as such ideal for karaoke games that necessitate the compilation of the music according to the respective notes.

Current Hits

The software requires only a one time processing of about 15 minutes for the compilation of songs for your games use. The recognition of the melody sung by the player during the game is conducted in real time with a minimal latency of only 6 ms.

Comprehensive Database

The minimal processing time per title and the automated detection makes it possible for you to compile an extensive database for melody tunes within an extremely short time.

Reliable Detection

The Fraunhofer technology is very robust against noise interference and the detection of the melodic pitch works even under difficult acoustic conditions. For the third time, the technology has been awarded the first prize by the MIREX international competition, in the category “Audio Melody Extraction”.

Cost and Resource Savings

Benefit from the Fraunhofer as an innovation leader, and save on cost and resource intensive in-house developments. The Fraunhofer Melody Extraction is shipped as a ready-to-use, out-of-the-box solution and also has very minimal impact on the processor load during operation.


The melody-detection is based on the automated transcription of music. The computer hereby attempts the opposite of making music. Instead of playing a music piece according to the score-sheet, the notes are detected from a music recording. Of course, such a transcription can also be carried out by experienced musicians and musicologists – but this takes time and requires high concentration. A "normal" music listener can rely on the help of the computer that is perhaps not as accurate as a musically trained ear, but much faster. The transcription of the main melody is particularly difficult. Often one first has to identify the most important voice among several voices. The so-called MRFFT-algorithm of the Fraunhofer IDMT is particularly suitable for this difficult task and has therefore for the third consecutive year been awarded the first place in the international MIREX competition in the category "Audio Melody Extraction".

For more information on the MRFFT-algorithm of the Fraunhofer IDMT, please see the white paper "Audio Melody Extraction for MIREX 2009".


The Melody Extraction for Music Games is available as C++ Library for Windows, Mac OS and Linux.

Test software

You would like to test the Melody Extraction? No problem. A free test version is available on demand.

This might also be of interest


Music learning software based on automatic music transcription

Research topic

Audio and Visual Content Analysis

Extracting meaningful data from audiovisual content


Research topic

Automatic Music Analysis