#generative-ai

24 Jul 2026 · Study
Generative AI · Text-to-speech
Text-to-speech with a small end-to-end model (Meta's MMS-TTS English, a VITS model): text to waveform in one network, plus a look at the generated audio's waveform and spectrogram.
16 Jul 2026 · Study
Generative AI · Text-to-image
How text steers a diffusion model: conditioning, classifier-free guidance, and the guidance scale trade-off, shown with a small conditional diffusion model on 2-D data where "prompts" are cluster labels.
1 Jul 2026 · Study
Generative AI · Stable Diffusion
Latent diffusion runs the diffusion process in a compressed latent space from a VAE; encoding and decoding a real photo with the Stable Diffusion VAE to see the 48× compression the whole model relies on.
23 Jun 2026 · Study
Generative AI · Diffusion models
DDPM from scratch on 2-D data: a fixed forward process that adds Gaussian noise, a small network trained to predict that noise, and a reverse loop that turns pure noise back into data.
16 Jun 2026 · Study
Generative AI · DCGAN
The DCGAN architecture guidelines (strided convolutions, batch norm, ReLU/LeakyReLU, no pooling or fully connected hidden layers) built in PyTorch, with the tensor shapes traced from noise to a 32×32 image.
8 Jun 2026 · Study
Generative AI · GANs
A generator and a discriminator trained against each other; the minimax game written out, a small GAN learning a 2-D ring of Gaussians, and the mode-collapse failure it can fall into.
1 Jun 2026 · Study
Generative AI · Variational autoencoders
A VAE learns a smooth latent space by encoding inputs as distributions and decoding samples from them; a small VAE on handwritten digits, the reparameterisation trick, and generating new digits from noise.
28 Mar 2026 · Projects
Built this: Modern DCGAN reproducibility study
A DCGAN (Radford et al., 2015) reproduction that keeps the all-convolutional architecture but swaps in LSGAN loss, spectral normalisation and label smoothing, plus an analysis toolkit for filters and feature maps.