List your company

Text-to-Speech on Japanese ambiguous pronunciation dataset (Male)

95.8AccuracyTKTO
48.16860.53472.985.266Oct 7, 2025

Evaluation results

MethodLinks
TKTOPO Data=Unpaired2025.1095.88.52.7
KTOPO Data=Unpaired2025.1095.28.73.2
gpt-4o-mini-tts2025.1093.911.16.2
gemini-2.5-pro-preview-tts2025.1088.511.97.3
gemini-2.5-flash-preview-tts2025.1076.913.49.1
TKTOPO Data=Paired2025.1070.16.62.9
DPOPO Data=Paired2025.1069.3138.2
Base model2025.1066.813.89.5
Supervised Fine-Tuning (SFT)PO Data=Desirable2025.1065.4138.4
KTOPO Data=Paired2025.1065.17.43
F5-TTS2025.105017.718.3
F5-TTS with G2P2025.105014.610.7