Glossary of AI video and image generation
83 terms, each defined where it actually matters on a production rather than in the abstract, and each linked to the explainer it comes from.
3
3D VAE
A video autoencoder that compresses along time as well as height and width.
A
Accept rate
The proportion of generations a human keeps, which is the number that actually determines cost per finished shot.
Appearance-free condition
A control signal such as a camera pose or trajectory chosen so it alters motion without changing the shape or look of objects.
attention
An operation where each position pulls a weighted mixture of information from all other positions.
B
Base checkpoint
The specific pretrained model an adapter was trained against and must be loaded with.
Blind super-resolution
Upscaling when the degradation that damaged the source is unknown, trained on synthetic degradation pipelines instead.
C
Cadence
The rhythm of motion set by frame rate and shutter, which stops matching the picture when frames are synthesised after the fact.
Camera trajectory conditioning
Supplying a sequence of camera poses as an explicit input so the generated clip approximates a specified move.
Case-based generalisation
Producing an answer by imitating the closest training example instead of applying an abstracted rule.
channels
The many values stored per latent cell, describing content rather than colour.
Character sheet
A prepared set of reference frames covering angles, expressions, wardrobe and lighting, used as conditioning across a sequence.
checkerboard artifact
Regular grid patterning produced by certain upsampling layers in a decoder.
classifier-free guidance
Running the model with and without conditioning and pushing away from the unconditional prediction to strengthen prompt adherence.
Community licence
A vendor-written licence that permits broad use with carve-outs, such as territorial exclusions or separate commercial terms, and is not an open-source licence.
compression ratio
How many pixels of input each unit of latent stands for, trading detail for speed and length.
Conditioning
Any input other than the initial noise that biases what the model generates, including text, images, masks, structure maps and camera trajectories.
Conditioning scale
A weight that sets how strongly a structure control overrides the model's own tendencies, traded against image quality.
Context length
The maximum number of tokens a model can attend over at once, which sets the longest clip it can generate in one pass.
contrastive training
Learning by matching images to their captions among many wrong options, as CLIP does.
Cross-attention
The layer type that lets a diffusion model read a conditioning signal such as an encoded text prompt while denoising.
D
decoder
The half that turns a finished latent back into pixels, run once at the end of generation.
Denoising steps
The number of iterative passes over the whole latent clip, each one a full-cost forward pass.
Detail hallucination
Invented high-frequency content that looks correct in isolation but does not correspond to anything in the source.
Determinism
The property that identical inputs produce identical outputs, which in practice holds only within a fixed software, hardware and model configuration.
distillation
Training a student model to reproduce a many-step sampler's output in far fewer steps.
DiT
Diffusion Transformer: a latent diffusion model whose denoising backbone is a transformer rather than a U-Net.
E
encoder
The half of the autoencoder that turns pixels into a latent, used during training and for image-to-video conditioning.
F
Fair dealing
The narrower UK and Commonwealth equivalent, available only for specified purposes rather than as an open-ended balancing test.
Fair use
A US statutory defence that can excuse otherwise infringing copying based on a multi-factor analysis, applied case by case rather than as a blanket rule.
forward process
The fixed schedule of adding noise to a real image, used to manufacture training examples.
Frame interpolation
Synthesising frames between existing ones, usually by estimating intermediate optical flow, to raise frame rate or slow motion.
frozen encoder
A text encoder whose weights stay fixed while the diffusion model trains around it.
G
Generation log
The record of prompt, seed, model version, parameters and conditioning asset hashes needed to reconstruct a delivered frame.
Generator
The stateful random-number object a pipeline draws from, which must be re-seeded per call to repeat a result.
guidance scale
How strongly that push is applied, trading prompt fidelity against variety.
I
Identity adapter
A small trained module that lets a frozen base model read a reference image of a subject as a prompt, IP-Adapter being the common example.
identity drift
A face or character gradually changing across a shot because the model stores no persistent identity.
Indemnification
A vendor's contractual promise to cover defined legal costs arising from use of its outputs, usually subject to caps, conditions and correct use of the product.
J
joint spatial-temporal attention
Attention computed across all positions in space and time at once, rather than separately per frame.
K
keyframe cascade
Generating distant frames first and then filling the gaps with temporal super-resolution models.
L
latent
A compressed, learned representation of an image or clip that the diffusion model operates on instead of pixels.
latent diffusion
Running the denoising process inside a pretrained autoencoder's compressed space rather than in pixel space.
LoRA
Low-rank adaptation: a training method that freezes the base weights and learns small rank-decomposition matrices inserted into each layer.
M
Motion prior
The statistical tendency a model has learned about how images move, which fills in everything the prompt and conditioning do not specify.
Multidistrict litigation
A US procedure that consolidates related federal cases before one judge for pretrial purposes, which is why several separate complaints can appear under a single docket.
N
noise prediction
The model's actual output at each step: an estimate of the noise in the current input, not the finished image.
O
Object permanence
Whether things persist correctly through occlusion and after leaving frame, which current models hold often but not reliably.
Object state change
The correct transformation of a thing under an action, such as food being eaten, which the Sora report lists as unreliable.
On-premises inference
Running generation on hardware you control so confidential material never leaves your network.
Open Source AI Definition
The Open Source Initiative's 1.0 standard, which asks for information about training data, the training and inference code, and the weights.
Open weights
A model whose trained parameters can be downloaded and run locally, regardless of whether its licence meets any open-source standard.
Out-of-distribution generalisation
Performance on cases unlike anything in training, where video models have been shown to fail rather than extrapolate a rule.
Outpainting
Generating new content beyond the existing frame edges, used to extend a plate or change aspect ratio.
Overfitting
Training so far into the dataset that the model reproduces its incidental details and loses responsiveness to prompts.
P
Parallax
The differential apparent motion of near and far objects under a moving viewpoint, which a generative model imitates rather than computes.
patch
A fixed-size square of latent treated as one token by the transformer.
Prior preservation
A training term that keeps a fine-tune from overwriting the model's general knowledge of the class your subject belongs to.
Provenance record
The log of model, model version, prompt, seed, parameters and input references behind a generated shot, kept so a delivery can be explained later.
R
Rank
The size of the low-rank update, which sets how much the adapter can learn and how large the resulting file is.
reconstruction artifact
A defect introduced by decoding rather than by the diffusion process, so it repeats across seeds.
reverse process
The learned sequence of denoising steps that turns pure noise into a sample.
S
sample
One generation drawn from the model, which differs run to run because the starting noise differs.
scaling behaviour
The observed relationship between compute or model size and output quality, which lets teams predict the return on more spending.
scheduler
The rule that decides step sizes and how each noise prediction is applied; also called the sampler.
score
The direction in which a noisy image should move to become more like real data; the quantity score-based formulations estimate.
Seed
The integer that initialises the pseudo-random number generator producing the starting noise for a generation.
sequence length
How many tokens one generation contains, growing with resolution and duration.
Spacetime patch
The unit a video transformer processes, a small block of latent covering an area of frame across a span of frames.
step
One pass of the model plus one scheduler update, moving the latent partway from noise toward a picture.
Structure conditioning
Conditioning on a spatial map derived from a reference, such as depth, edges or human pose, to fix geometry while leaving appearance to the model.
Subject fine-tuning
Updating model weights, or an adapter on them, from a small image set so a specific person, object or set can be called up by name.
Super-resolution
Increasing pixel dimensions by predicting the detail a higher-resolution version of the image would plausibly contain.
T
temporal flicker
Frame-to-frame instability in texture, brightness or fine detail that was never committed to in the latent.
Temporal flickering
Frame-to-frame instability in texture, colour or detail, tracked as its own evaluation dimension in VBench.
temporal layers
Blocks added to a pretrained image model so it can relate frames to each other, an alternative to dicing along time.
temporal super-resolution
A separately trained model that increases frame rate by generating intermediate frames.
text encoder
A separate network that converts prompt tokens into vectors the diffusion model can attend to.
Textual Inversion
Learning a new embedding, effectively a new word, for a concept while leaving the model itself unchanged.
texture crawl
Patterns and fine surfaces reshuffling each frame instead of holding still on the object.
token
One position the model attends over; in image and video models, one patch of latent.
Trigger word
The token a fine-tune binds a subject or style to, used to call that concept up in a prompt.
U
U-Net
The convolutional encoder-decoder backbone used by earlier diffusion models, which processes the frame through a hierarchy of resolutions.
V
Version pinning
Locking a production to one exact model build so results stay stable across a revision cycle.