Text-to-Speech on German unseen languages
6.21WERGround Truth
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| Ground TruthInference Mode=Zero-shot evaluation, Training Steps=300K steps2026.06 | 6.21 | — | 2.481 | 2.18 | |
| Pixel-TTSInference Mode=Zero-shot evaluation, Training Steps=300K steps2026.06 | 66.48 | 48.1 | 3.303 | 27.36 | |
| Text-TTSInference Mode=Zero-shot evaluation, Training Steps=300K steps2026.06 | 71.49 | 48.1 | 3.29 | 31.5 |