Long-form Speech Generation on HiFiTTS long-form 1-hour
2.5WERMagpieTTS-LF
Evaluation Results
| Method | Links | ||||
|---|---|---|---|---|---|
| MagpieTTS-LFeps=0.1, w=(0.2, 0.8, 1.0, 0.8, 0.2), temperature=0.7, lambda=1.0, cfg scale=2.52026.06 | 2.5 | 1.2 | 79 | 97.9 | |
| Qwen3-TTSselection_criteria=best performing configuration2026.06 | 4.5 | 2.8 | 80 | 95.8 | |
| XTTS2026.06 | 5.1 | 3.5 | 69 | 92.9 | |
| VibeVoiceselection_criteria=best performing configuration2026.06 | 11.5 | 10.5 | 53 | 84.8 |