Vocos: Closing the gap between time-domain and Fourier-based neural vocoders for high-quality audio synthesis.
Hubert SiuzdakPublished in: ICLR (2024)
Keyphrases
- high quality
- frequency domain
- fourier transform
- network architecture
- multimedia
- neural network
- low quality
- fourier spectrum
- ground truth
- audio visual
- bio inspired
- image reconstruction
- visual data
- radon transform
- digital video
- program synthesis
- series expansion
- audio signals
- emotion recognition
- visual information
- neural model
- fourier domain
- fourier analysis
- signal processing
- image processing