Research map / Image generation / GANs
Adversarial Audio Synthesis
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 phase shuffle n=2 to WaveGAN (ablation) | WaveGAN · SC09 | Inception score: 4.12 ± 0.03 -> 4.67 ± 0.01 | Helped |
| increase phase shuffle from n=2 to n=4 (ablation) | WaveGAN · SC09 | Inception score: 4.67 ± 0.01 -> 4.54 ± 0.03 | Hurt |
| replace transposed convolution with nearest-neighbor upsampling (ablation) | WaveGAN · SC09 | Inception score: 4.67 ± 0.01 -> 3.77 ± 0.02 | Hurt |
| add a learned post-processing filter (ablation) | WaveGAN · SC09 | Inception score: 4.67 ± 0.01 -> 3.92 ± 0.03 | Hurt |
| add 50% dropout in discriminator activations (ablation) | WaveGAN · SC09 | Inception score: 4.67 ± 0.01 -> 3.93 ± 0.03 | Hurt |
| add phase shuffle n=1 (ablation) | SpecGAN · SC09 | Inception score: 6.03 ± 0.04 -> 3.71 ± 0.03 | Hurt |
| generate spectrograms instead of raw waveforms | WaveGAN and SpecGAN · SC09 | Inception score: 4.7 vs. 6.0 | Helped |
| generate raw waveforms instead of spectrograms | WaveGAN and SpecGAN · SC09 | Sound quality: 2.3 ± 0.9 vs. 1.9 ± 0.8 | Helped |
| generate raw waveforms instead of spectrograms | WaveGAN and SpecGAN · SC09 | Speaker diversity: 3.2 ± 0.9 vs. 2.6 ± 1.0 | Helped |
Closest papers by meaning
- GANSynth: Adversarial Neural Audio Synthesis (ICLR 2019 · similarity 0.93)
- MelGAN: Generative Adversarial Networks for Conditional Waveform Synthesis (NeurIPS 2019 · similarity 0.91)
- Catch-A-Waveform: Learning to Generate Audio from a Single Short Example (NeurIPS 2021 · similarity 0.90)
- High Fidelity Speech Synthesis with Adversarial Networks (ICLR 2020 · similarity 0.90)
- BigVGAN: A Universal Neural Vocoder with Large-Scale Training (ICLR 2023 · similarity 0.89)
- HiFi-GAN: Generative Adversarial Networks for Efficient and High Fidelity Speech Synthesis (NeurIPS 2020 · similarity 0.89)
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.