GANStrument: Adversarial Instrument Sound Synthesis with Pitch-invariant Instance Conditioning

TL;DR

GANStrument is a novel adversarial model that synthesizes realistic instrument sounds from a single input, maintaining timbre and pitch invariance, with improved fidelity, diversity, and input flexibility.

Contribution

It introduces a pitch-invariant feature extractor and instance conditioning to enhance sound synthesis quality and generalization in GAN-based models.

Findings

Outperforms baseline models in sound quality and diversity

Achieves better pitch accuracy and timbre consistency

Demonstrates improved input editability and generalization

Abstract

We propose GANStrument, a generative adversarial model for instrument sound synthesis. Given a one-shot sound as input, it is able to generate pitched instrument sounds that reflect the timbre of the input within an interactive time. By exploiting instance conditioning, GANStrument achieves better fidelity and diversity of synthesized sounds and generalization ability to various inputs. In addition, we introduce an adversarial training scheme for a pitch-invariant feature extractor that significantly improves the pitch accuracy and timbre consistency. Experimental results show that GANStrument outperforms strong baselines that do not use instance conditioning in terms of generation quality and input editability. Qualitative examples are available online.