An interactive tool for voice transformation with a neural auto-encoder.
Circe lets you modify the pitch and tension of the voice. Used as a plug-in, these parameters can be controlled automatically by analyzing these characteristics from a second audio track defined as a side-chain. Circe can also be used to create a hybrid timbre between the original track and the one defined as a side-chain.
Circe is available as a VST3 and AudioUnit plug-in and as a standalone application that can be used as an external sample editor. Circe is compatible with Windows, Linux, and MacOS (M series only).
Circe neural auto-encoder model is developed by Frederik Bous and Axel Roebel at IRCAM Analysis-Synthesis Team of the STMS Laboratory.
Circe plug-in and Circe standalone application are designed and developed by Pierre Guillot at IRCAM IMR department.
https://forum.ircam.fr/projects/detail/circe/
