Sample display: Prompt → Generated text → Clean and noisy audio at 6 / 0 / -6 dB.
| Gen Text | Prompt Audio | Noise | 6 dB (clean / noisy) | 0 dB (clean / noisy) | -6 dB (clean / noisy) |
|---|---|---|---|---|---|
正在加载 demo_manifest.csv ... |
|||||
Comparing the proposed NIL-TTS model with baseline models under identical conditions.
| Gen Text | Prompt Audio | Noise | F5-TTS (Baseline) (Clean / Noisy) |
NIL-TTS (Proposed) (Clean / Noisy) |
MetaVoice-1B Lomb (Clean / Noisy) |
F5TTS-Style Lomb (Clean / Noisy) |
RL-only (Clean / Noisy) |
SFT-only (Clean / Noisy) |
|---|---|---|---|---|---|---|---|---|
正在加载 compare_manifest.csv ... |
||||||||
Comparing matched vs. mismatched speech synthesized for different environments, mixed under the same noise.
| Gen Text | Prompt Audio | Conditioning Noise | Gen Audio (Clean) |
Mixed Audio under Same Noise (Match vs. Mismatch) |
|---|---|---|---|---|
正在加载 mismatch_manifest.csv ... |
||||