RAVE: A variational autoencoder for fast and high-quality neural audio synthesis.
Antoine CaillonPhilippe EslingPublished in: CoRR (2021)
Keyphrases
- high quality
- network architecture
- neural network
- multimedia
- program synthesis
- image segmentation
- low quality
- signal processing
- audio video
- ground truth
- bio inspired
- audio visual
- high resolution
- neural model
- higher quality
- optical flow
- biologically inspired
- associative memory
- visual information
- cross modal
- audio signals
- depth map
- methods in computer vision
- visual data
- variational framework
- free energy
- restricted boltzmann machine
- hebbian learning
- music score
- audio stream