Autoencoder
Definition
A neural network that compresses an input into a compact latent representation (encoder) then reconstructs it (decoder). Face-swap pipelines train autoencoders with a shared encoder but separate decoders for each identity, enabling identity transplantation.
- Structure
- Encoder plus decoder
- Output of encoder
- Compact latent representation
- Face-swap design
- Shared encoder, separate decoder per identity
Common questions
Why do face-swap pipelines share one encoder across two identities?+
Sharing the encoder forces the latent space to capture identity-neutral structure such as pose and expression, so swapping which decoder reconstructs from that latent transplants only the identity, not the pose.
Is an autoencoder the same architecture as a GAN?+
No. A GAN trains a generator against a discriminator in an adversarial setup, while an autoencoder trains by minimizing reconstruction error. Some deepfake pipelines combine both ideas in a single model.
Related terms
- Blending Mask
- In face-swap pipelines, a pixel-level mask that defines the face region to be composited onto the target frame. Imperfect masks leave boundary...
- Checkerboard Artifact
- A grid-like pattern visible in the Fourier power spectrum of GAN outputs, caused by transposed convolution or bilinear-upsampling operations used to increase...
- Diffusion Model
- A generative neural network architecture (Ho et al., 2020; Stable Diffusion, Rombach et al., 2022) that learns to reverse a noise-addition process...
- Generative Adversarial Network (GAN)
- A machine learning architecture consisting of two networks trained simultaneously: a generator that produces synthetic images and a discriminator that classifies images...
- Latent Space
- The compressed, lower-dimensional representation of data learned by a neural network's internal layers. Generative models sample from or navigate this space to...