Root cause of F5-TTS’s anomalous Japanese intelligibility

Ascertain the root cause of F5-TTS’s near-complete intelligibility failure on Japanese observed in the reported evaluation, including why the Japanese character error rate exceeded 1.0, by analyzing F5-TTS’s internal processing or equivalent diagnostic evidence.

Background

In the comparative evaluation, F5-TTS exhibited a Japanese character error rate greater than 1.0, indicating severe intelligibility failure despite competitive performance in other conditions.

The authors state they cannot determine the reason without access to F5-TTS’s internal processing, explicitly marking the cause as unclear and outside the scope of their study.

References

The root cause is unclear without access to F5-TTS's internal processing, which is beyond the scope of this study.

T5Gemma-TTS Technical Report  (2604.01760 - Arata et al., 2 Apr 2026) in Results, Section 5.1 (In-Training-Language Evaluation)