Synthesizing the Lombard Effect: Multi-Level Control of Speech Clarity and Vocal Effort in TTS

The article presents a flow-matching based text-to-speech (TTS) model designed to simulate the Lombard effect, enabling continuous control over vocal effort and articulation. The model incorporates pseudo-labels for these parameters and allows for word-level emphasis, which enhances clarity in challenging acoustic environments. Experimental results indicate that the model effectively improves speech intelligibility in noisy conditions, making it significant for practitioners focused on developing more adaptive and human-like TTS systems.