Granular synthesis is a flexible method to create a wide range of complex sounds, like the sound of rain or water, using very short waveforms, called grains. To synthesize realistic, natural sounds appropriate grains are needed. In an earlier paper we already presented a method to extract grains from recordings of complex sounds. In this paper we describe an extension of the earlier method in which the end of incomplete grains is estimated to improve sound quality. Additionally synthesis parameters that allow us to recreate sound output very close to the original recordings are found automatically. A few seconds of audio input will provide enough data to synthesize sounds of arbitrary length. The necessary grains only require little memory and since synthesis parameters can also be varied to change the nature of the sound, this method is especially beneficial for video games. While empirical listening suggests that the synthesized waveforms sound natural, a formal listening test was not conducted. Sound samples are provided.
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.