Skip to content

Releases: primepake/learnable-speech

DAC-VAE 24kHz Mono - Continuous Latent Space Model

03 Jul 16:37

Choose a tag to compare

What's New

This release introduces the DAC-VAE model configured for 24kHz mono audio with continuous latent space representation.

Key Features

  • Sample Rate: 24kHz mono audio processing
  • Architecture: Variational Autoencoder (VAE) with continuous latent space
  • Model Type: Discrete Audio Codec (DAC) based VAE implementation

Technical Details

  • Optimized for single-channel (mono) audio at 24kHz sampling rate
  • Continuous latent space representation for smooth interpolation between audio samples
  • Suitable for audio generation, compression, and manipulation tasks

Usage

cd dac-vae
python inference.py --checkpoint checkpoint.pt --config config.yml