SupertonicTTS: Towards Highly Efficient and Streamlined Text-to-Speech System
This work addresses efficiency and complexity issues in TTS systems for applications requiring lightweight models, though it appears incremental as it builds on existing TTS paradigms.
The authors tackled the problem of inefficient and complex text-to-speech systems by introducing SupertonicTTS, which achieves performance comparable to zero-shot TTS models with only 44M parameters while reducing architectural complexity and computational cost.
We introduce SupertonicTTS, a novel text-to-speech (TTS) system designed for efficient and streamlined speech synthesis. SupertonicTTS comprises three components: a speech autoencoder for continuous latent representation, a text-to-latent module leveraging flow-matching for text-to-latent mapping, and an utterance-level duration predictor. To enable a lightweight architecture, we employ a low-dimensional latent space, temporal compression of latents, and ConvNeXt blocks. The TTS pipeline is further simplified by operating directly on raw character-level text and employing cross-attention for text-speech alignment, thus eliminating the need for grapheme-to-phoneme (G2P) modules and external aligners. In addition, we propose context-sharing batch expansion that accelerates loss convergence and stabilizes text-speech alignment with minimal memory and I/O overhead. Experimental results demonstrate that SupertonicTTS delivers performance comparable to contemporary zero-shot TTS models with only 44M parameters, while significantly reducing architectural complexity and computational cost. Audio samples are available at: https://supertonictts.github.io/.