Research map / Image generation / Autoregressive and latent generation
End-to-end Adversarial Text-to-Speech
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add adversarial discriminators (ablation) | EATS · held-out conditioning sequences | MOS: 1.407 ± 0.040 -> 4.083 ± 0.049 | Helped |
| add random window discriminators (ablation) | EATS · held-out conditioning sequences | MOS: 2.526 ± 0.060 -> 4.083 ± 0.049 | Helped |
| use phoneme input instead of character input (ablation) | EATS · held-out conditioning sequences | MOS: 3.423 ± 0.073 -> 4.083 ± 0.049 | Helped |
| add the mel-spectrogram discriminator (ablation) | EATS · held-out conditioning sequences | MOS: 3.525 ± 0.057 -> 4.083 ± 0.049 | Helped |
| use monotonic interpolation instead of transformer attention (ablation) | EATS · held-out conditioning sequences | MOS: 3.551 ± 0.073 -> 4.083 ± 0.049 | Helped |
| add dynamic time warping to the prediction loss (ablation) | EATS · held-out conditioning sequences | MOS: 3.559 ± 0.065 -> 4.083 ± 0.049 | Helped |
| train on multispeaker data instead of single-speaker data (ablation) | EATS · held-out conditioning sequences for the evaluated speaker | MOS: 3.829 ± 0.055 -> 4.083 ± 0.049 | Helped |
| add the length prediction loss (ablation) | EATS · training | training convergence | Helped |
| add the spectrogram prediction loss (ablation) | EATS · training | training convergence | Helped |
Closest papers by meaning
- Conditional Variational Autoencoder with Adversarial Learning for End-to-End Text-to-Speech (ICML 2021 · similarity 0.90)
- High Fidelity Speech Synthesis with Adversarial Networks (ICLR 2020 · similarity 0.90)
- Multi-SpectroGAN: High-Diversity and High-Fidelity Spectrogram Generation with Adversarial Style Combination for Speech Synthesis (AAAI 2021 · similarity 0.88)
- Transfer Learning from Speaker Verification to Multispeaker Text-To-Speech Synthesis (NeurIPS 2018 · similarity 0.88)
- Advancing Audio-Based Text Generation with Imbalance Preference Optimization (AAAI 2025 · similarity 0.87)
- NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers (ICLR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.