Suno Controls: Weirdness, Style Influence, and Audio Influence

This paper presents a corrected and behaviorally grounded framework for understanding the three core controls in generative audio systems: Weirdness, Style Influence, and Audio Influence. Rather than treating these controls as artistic labels or surface parameters, the paper defines them functionally as interacting behavioral forces that shape system decision-making in real time. Weirdness is defined as the system’s open decision space, governing the degree of improvisation and exploratory behavior. Style Influence is redefined as directive linguistic pressure, analogous to a conductor imposing an interpretive framework, rather than a tool for pronunciation or quality correction. Audio Influence is clarified as the degree of system intervention in the existing performance, determining whether the system adheres to or reconstructs the provided audio. Special emphasis is placed on the zero-value behavior of each control, demonstrating that misinterpretation at these boundary conditions leads to systematic errors, degraded articulation, and unstable outputs. The paper further establishes that common audio defects—such as distortion, swallowed phonemes, loss of continuity, and lifeless performance—are not caused by individual controls, but by imbalance and role confusion between them. The framework concludes that quality in generative audio does not emerge from increased control, but from a clear separation of roles between decision freedom, directive pressure, and system intervention. This corrected model replaces earlier interpretive assumptions and provides a stable, operational reference for both practitioners and researchers working with probabilistic generative audio systems.