Accented Speech Synthesis on LibriTTS-R (train-clean-100)
73.8US ProbabilityKokoro-82M (US Speaker Embedding)
Evaluation results
| Method | Links | |||||||
|---|---|---|---|---|---|---|---|---|
| Kokoro-82M (US Speaker Embedding)Speaker Embedding=US, Phonological Rules=None2026.03 | 73.8 | — | 0.71 | — | 4.38 | 3.42 | 2.83 | |
| Kokoro-82M (Spanish Speaker Embedding)Speaker Embedding=Spanish, Phonological Rules=None2026.03 | 26.6 | 23.7 | 0.73 | 47 | 3.73 | 7.99 | 5.64 | |
| Kokoro-82M (US Speaker Embedding) + Spanish RulesSpeaker Embedding=US, Phonological Rules=Spanish2026.03 | 25.97 | — | 0.5 | — | 4.39 | 24.91 | 13.56 | |
| Kokoro-82M (US Speaker Embedding) + Indian RulesSpeaker Embedding=US, Phonological Rules=Indian2026.03 | 16.85 | — | 0.42 | — | 4.38 | 32.66 | 18.32 | |
| Kokoro-82M (Spanish Speaker Embedding) + Spanish RulesSpeaker Embedding=Spanish, Phonological Rules=Spanish2026.03 | 2.97 | 51.59 | 0.38 | 60 | 3.87 | 13.13 | 7.79 | |
| Kokoro-82M (Indian Speaker Embedding)Speaker Embedding=Indian, Phonological Rules=None2026.03 | 1.89 | 58.86 | 0.28 | 74 | 4.25 | 8.35 | 5.68 | |
| Kokoro-82M (Indian Speaker Embedding) + Indian RulesSpeaker Embedding=Indian, Phonological Rules=Indian2026.03 | 0.27 | 86.4 | -0.008 | 75 | 4.16 | 24.15 | 14.78 |