Sarashina2.2-TTS: Tackling Kanji Polyphony in Japanese Speech Generation via Data Scaling and Targeted Data Synthesis

2026-06-23

AI Summary

This work addresses the challenge of context-dependent pronunciation of kanji homographs in Japanese text-to-speech synthesis by proposing a large language model–based TTS system. Trained on 361k hours of multilingual data and augmented with targeted data covering all 2,136 Jōyō kanji (encompassing 4,378 distinct readings), the system introduces the Jōyō Kanji Yomi Benchmark for evaluation and employs a novel Kana-CER metric to directly assess pronunciation accuracy in kana space. Experimental results demonstrate state-of-the-art performance in character-level reading accuracy, sentence-level prosody comparable to top-tier baselines, superior speaker similarity in zero-shot Japanese synthesis, and robust pronunciation consistency across arbitrary prompt languages.

Abstract

While large language model (LLM)-based text-to-speech (TTS) systems have achieved high-quality speech synthesis, most existing systems focus on English and Chinese. Japanese, however, remains under-explored, and its unique linguistic challenges, such as widespread context-dependent kanji polyphony, have yet to be adequately tackled. Here we introduce Sarashina2.2-TTS (https://github.com/sbintuitions/sarashina2.2-tts), a Japanese-centric LLM-TTS system that tackles these challenges through a dual approach: data strategy and evaluation methodology. First, we scale training to approximately 361k hours of speech, incorporating a balanced mix of Japanese and English data. Furthermore, we design a targeted data augmentation pipeline covering all 2,136 Joyo (regular-use) kanji designated by Japan's Agency for Cultural Affairs to efficiently address kanji polyphony disambiguation. Second, we introduce the Joyo Kanji Yomi Benchmark (https://github.com/sbintuitions/JoyoKanji-Yomi-Benchmark), covering all 2,136 Joyo kanji and their 4,378 readings. Alongside this benchmark, we propose Kana-CER, a metric that compares synthesized speech against reference readings in the kana space, eliminating orthographic variations to directly measure pronunciation correctness. Experiments demonstrate that our targeted data augmentation significantly improves reading accuracy. Overall, Sarashina2.2-TTS achieves state-of-the-art kanji-level reading accuracy and matches top baselines on general sentence-level pronunciation, while delivering the highest speaker similarity in zero-shot Japanese speech synthesis. Furthermore, cross-lingual evaluation reveals that Sarashina2.2-TTS is the only system that maintains stable Japanese pronunciation regardless of the prompt language, confirming that our balanced training approach improves cross-lingual robustness.

Problem

• kanji polyphony
• Japanese speech synthesis
• pronunciation disambiguation
• text-to-speech
• multilingual TTS

Innovation

• kanji polyphony
• targeted data synthesis
• Joyo Kanji Yomi Benchmark
• Kana-CER
• cross-lingual TTS