Free Information Center
Diffsound (discrete diffusion for audio) is a method in generative audio modeling that applies discrete diffusion processes to create or transform audio data. It is part of a broader class of diffusion-based generative models adapted specifically for the discrete, sequential nature of audio signals.
A latent diffusion model (LDM) is a type of generative machine learning model that performs diffusion processes in a compressed latent space, enabling efficient and high-quality image synthesis and related tasks. By operating in a lower-dimensional representation, LDMs reduce computational costs while maintaining detailed output.
Denoising diffusion probabilistic models (DDPM) are a class of generative models that produce data by reversing a gradual noising process. They have gained prominence for their ability to generate high-quality images and other complex data.
The denoising diffusion implicit model (DDIM) is a generative modeling technique that improves the efficiency and sampling speed of diffusion-based models by introducing a non-Markovian diffusion process. It enables faster image synthesis while maintaining high-quality results.
DreamFusion is a text-to-3D synthesis method that generates three-dimensional models from textual descriptions using neural rendering and diffusion models. It enables the creation of 3D objects without requiring explicit 3D data inputs.
WaveGrad is a diffusion-based vocoder that generates speech waveforms from mel-spectrograms using a generative diffusion probabilistic model. It offers an alternative to traditional neural vocoders by progressively refining noise into audio, achieving high-quality speech synthesis.
DiffWave is a generative model based on diffusion processes for high-quality waveform synthesis, primarily used in speech generation. It leverages a denoising diffusion probabilistic model to produce natural audio waveforms from noise, offering an alternative to traditional autoregressive and adversarial approaches.
A score-based generative model is a type of probabilistic model that generates data by estimating the gradient of the data distribution’s log-density, known as the score function. These models employ techniques from stochastic differential equations and denoising score matching to synthesize high-quality samples, especially in image generation.