New AI model enables continuous control of vocal intensity in speech

Researchers have developed a novel method for vocal intensity conversion that allows for continuous control over sound pressure level (SPL) using adversarial training. This approach moves beyond discrete categories or reference utterances, enabling users to specify a target SPL on a continuous dB scale. The system employs a lightweight converter, trained as a conditional cycle-consistent GAN, which modifies features from a frozen speech encoder. When using WavLM features, the converter successfully achieved 83% of the target intensity change and also influenced the fundamental frequency, with perceptual evaluations confirming increased perceived vocal intensity. AI

IMPACT This research could lead to more nuanced and controllable speech synthesis and manipulation tools.

RANK_REASON The item is an academic paper detailing a new method for speech processing. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.LG →

• arXiv
• CORE Recommender
• cycle-consistent GAN
• Hugging Face
• log-mel spectrograms
• SpeechTokenizer
• WavLM

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →