OZSpeech: One-step Zero-shot Speech Synthesis with Learned-Prior-Conditioned Flow Matching
This work addresses computational inefficiency and attribute modeling in speech synthesis, offering a more efficient and precise TTS system for applications requiring high-quality voice cloning.
The authors tackled the limitations of existing text-to-speech systems by introducing OZSpeech, a method that uses learned-prior-conditioned flow matching with one-step sampling to improve efficiency and accuracy, achieving promising performance in content accuracy, naturalness, prosody generation, and speaker style preservation.
Text-to-speech (TTS) systems have seen significant advancements in recent years, driven by improvements in deep learning and neural network architectures. Viewing the output speech as a data distribution, previous approaches often employ traditional speech representations, such as waveforms or spectrograms, within the Flow Matching framework. However, these methods have limitations, including overlooking various speech attributes and incurring high computational costs due to additional constraints introduced during training. To address these challenges, we introduce OZSpeech, the first TTS method to explore optimal transport conditional flow matching with one-step sampling and a learned prior as the condition, effectively disregarding preceding states and reducing the number of sampling steps. Our approach operates on disentangled, factorized components of speech in token format, enabling accurate modeling of each speech attribute, which enhances the TTS system's ability to precisely clone the prompt speech. Experimental results show that our method achieves promising performance over existing methods in content accuracy, naturalness, prosody generation, and speaker style preservation. Audio samples are available at our demo page https://ozspeech.github.io/OZSpeech_Web/.