Abstract: In recent years, the application of deep learning methods to guitar effect modelling has garnered considerable attention. Motivated by recent advancements in neural vocoders, we experimented with two different discriminator architectures in replacement of the discriminator employed in an existing GAN-based guitar amplifier modelling framework, for better audio quality. We report the preliminary result of evaluating these methods on the audio samples of 3 amplifiers from the EGDB dataset.