ARCHI-TTS: A flow-matching-based Text-to-Speech Model with Self-supervised Semantic Aligner and Accelerated Inference

TL;DR

ARCHI-TTS introduces a flow-matching-based TTS model with a semantic aligner and an efficient inference strategy, significantly improving synthesis quality and speed over existing diffusion-based TTS systems.

Contribution

It presents a novel semantic aligner and an inference acceleration method that together enhance robustness and efficiency in diffusion-based TTS.

Findings

Achieves 1.98% WER on LibriSpeech-PC test-clean

Outperforms recent state-of-the-art TTS systems in quality and speed

Reduces inference time by reusing encoder features

Abstract

Although diffusion-based, non-autoregressive text-to-speech (TTS) systems have demonstrated impressive zero-shot synthesis capabilities, their efficacy is still hindered by two key challenges: the difficulty of text-speech alignment modeling and the high computational overhead of the iterative denoising process. To address these limitations, we propose ARCHI-TTS that features a dedicated semantic aligner to ensure robust temporal and semantic consistency between text and audio. To overcome high computational inference costs, ARCHI-TTS employs an efficient inference strategy that reuses encoder features across denoising steps, drastically accelerating synthesis without performance degradation. An auxiliary CTC loss applied to the condition encoder further enhances the semantic understanding. Experimental results demonstrate that ARCHI-TTS achieves a WER of 1.98% on LibriSpeech-PC test-clean, and 1.47%/1.42% on SeedTTS test-en/test-zh with a high inference efficiency, consistently outperforming recent state-of-the-art TTS systems.