Research map / Image generation / GANs
High Fidelity Speech Synthesis with Adversarial Networks
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 |
|---|---|---|---|
| replace the full-input discriminator with random-window discriminators (ablation) | GAN-TTS · TTS human evaluation | MOS: 1.889 -> 4.213 | Helped |
| use multiple conditional random-window discriminators instead of a single conditional discriminator (ablation) | GAN-TTS · TTS human evaluation | MOS: 3.394 -> 4.213 | Helped |
| add unconditional random-window discriminators (ablation) | GAN-TTS · TTS human evaluation | MOS: 3.498 -> 4.213 | Helped |
| use a mixture of different discriminator window sizes rather than ten fixed-size windows (ablation) | GAN-TTS · TTS human evaluation | MOS: 3.526 -> 4.154 | Helped |
| use downsampling in the discriminators with different window sizes (ablation) | GAN-TTS · TTS human evaluation | MOS: 4.154 -> 4.213 | No clear effect |
| use a 240-sample rather than 480-sample base window (ablation) | GAN-TTS · TTS human evaluation | MOS: 4.195 -> 4.213 | No clear effect |
| apply the 16-bit μ-law transform to audio (ablation) | GAN-TTS · TTS generation | performance | Helped |
Closest papers by meaning
- HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis (NeurIPS 2020 · similarity 0.93)
- GANSynth: Adversarial Neural Audio Synthesis (ICLR 2019 · similarity 0.92)
- A Spectral Energy Distance for Parallel Speech Synthesis (NeurIPS 2020 · similarity 0.91)
- BigVGAN: A Universal Neural Vocoder with Large-Scale Training (ICLR 2023 · similarity 0.91)
- FINALLY: fast and universal speech enhancement with studio-like quality (NeurIPS 2024 · similarity 0.88)
- NaturalSpeech 2: Latent Diffusion Models are Natural and Zero-Shot Speech and Singing Synthesizers (ICLR 2024 · similarity 0.87)
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.