HyperGANStrument: Instrument Sound Synthesis and Editing with Pitch-Invariant Hypernetworks

TL;DR

HyperGANStrument tackles the challenge of high-fidelity instrument sound synthesis and editable timbre with a one-shot input by introducing a pitch-invariant hypernetwork that modulates a pre-trained GANStrument generator. The method combines a feedback-refinement loop and conditional adversarial fine-tuning to improve reconstruction fidelity, pitch accuracy, and timbre-pitch disentanglement. Experimental results on NSynth show clear gains over GANStrument and encoder-based inversion, including better realism, controllability, and generalization to non-instrument timbres, while maintaining practical generation times. This approach yields a lightweight, interactive neural sampler suitable for real-world music applications with diverse timbre editing capabilities.

Abstract

GANStrument, exploiting GANs with a pitch-invariant feature extractor and instance conditioning technique, has shown remarkable capabilities in synthesizing realistic instrument sounds. To further improve the reconstruction ability and pitch accuracy to enhance the editability of user-provided sound, we propose HyperGANStrument, which introduces a pitch-invariant hypernetwork to modulate the weights of a pre-trained GANStrument generator, given a one-shot sound as input. The hypernetwork modulation provides feedback for the generator in the reconstruction of the input sound. In addition, we take advantage of an adversarial fine-tuning scheme for the hypernetwork to improve the reconstruction fidelity and generation diversity of the generator. Experimental results show that the proposed model not only enhances the generation capability of GANStrument but also significantly improves the editability of synthesized sounds. Audio examples are available at the online demo page.

Figures (3)

• Figure 1: An overview of HyperGANStrument.
• Figure 2: Examples of reconstruction and generation.
• Figure 3: Examples of interpolation.