Sound synthesis represents an unavoidable tool for modern composers and performers, but achieving desired sounds often requires unintuitive control of numerical parameters. This doctoral thesis consists of three approaches to intuitive sound synthesis. The first approach is based on the selection of synthesis parameters in a way that generated sounds match input timbral attributes as closely as possible. The doctoral thesis presents three different methods for controlling synthesis parameters by using input attributes without a priori knowledge of underlying synthesizer, its parameters, or architecture. The second approach focuses on controlling synthesis parameters during live performances. The proposed method relies on following the current position of live performance in the referent score or in the recording and on applying parameter values at the appropriate moments. The third approach is specialized for controlling non-standard synthesis techniques and encompasses three modalities of controlling dynamic stochastic synthesis – using audio input, physical gestures, and MIDI protocol.The contributions of this PhD thesis are: (1) an algorithm for selection of sound synthesis parameters based on timbral attributes which does not depend on synthesis method or on synthesizer architecture, (2) parameter automation for live performances using score following with a modified dynamic time warping algorithm, and (3) methods for interactive control of dynamic stochastic synthesis using input audio signals, physical gestures, and MIDI messages.