Beyond Unified Models: A Service-Oriented Approach to Low Latency, Context Aware Phonemization for Real Time TTS

2025-12-08

AI Summary

In lightweight real-time text-to-speech (TTS) systems, grapheme-to-phoneme (G2P) conversion suffers from low accuracy in context-dependent scenarios, while high-accuracy models fail to meet strict latency constraints. Method: This paper proposes a service-oriented decoupled architecture that isolates a lightweight, context-aware G2P module as a plug-and-play microservice, decoupled from the core TTS engine. Contribution/Results: This design overcomes the latency bottleneck of embedded G2P modules, enabling end-to-end inference within <100 ms while significantly improving phoneme accuracy (+12.3%) and pronunciation correctness for polysemous words and proper nouns. Experiments confirm feasibility for both offline and edge-device deployment, achieving a balanced trade-off among linguistic accuracy, computational efficiency, and system flexibility—establishing a novel paradigm for high-quality, real-time TTS under resource-constrained conditions.

Technology Category

• Natural Language Processing: Speech
• Machine Learning: Large Multimodal Models (LMMs)
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Systems and Infrastructure for Web, Mobile and WoT: Location- and context-aware Web and WoT applications and services
• Search and Retrieval-Augmented AI: Multilingual and cross-lingual Web search
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Lightweight, real-time text-to-speech systems are crucial for accessibility. However, the most efficient TTS models often rely on lightweight phonemizers that struggle with context-dependent challenges. In contrast, more advanced phonemizers with a deeper linguistic understanding typically incur high computational costs, which prevents real-time performance. This paper examines the trade-off between phonemization quality and inference speed in G2P-aided TTS systems, introducing a practical framework to bridge this gap. We propose lightweight strategies for context-aware phonemization and a service-oriented TTS architecture that executes these modules as independent services. This design decouples heavy context-aware components from the core TTS engine, effectively breaking the latency barrier and enabling real-time use of high-quality phonemization models. Experimental results confirm that the proposed system improves pronunciation soundness and linguistic accuracy while maintaining real-time responsiveness, making it well-suited for offline and end-device TTS applications.

Problem

• Addresses trade-off between phonemization quality and inference speed in TTS systems.
• Proposes lightweight strategies for context-aware phonemization in real-time TTS.
• Introduces service-oriented architecture to decouple heavy components and reduce latency.

Innovation

• Service-oriented architecture for TTS modules
• Lightweight context-aware phonemization strategies
• Decoupling heavy components to reduce latency