47 TERMS,
NO FOG
Every term here is defined the way a producer needs it rather than the way a paper defines it: what it is, what it costs you when it goes wrong, and what to do about that. Each one has its own page with the questions people actually type.
- Terms
- 47
- Topics
- 7
- Written by
- People who ship the work
- Gate
- None
Agentic Workflow
A process where a model plans and executes several steps against real tools, with a person deciding at named checkpoints rather than approving every keystroke.
AI Slop
Generated content produced at volume without editorial judgement, recognisable by its smoothness, its symmetry and its complete absence of a point of view.
C2PA / Content Credentials
An open standard that attaches a signed, tamper-evident record of how a piece of media was made and edited, travelling with the file itself.
Camera Control
Directing a generative video model with the vocabulary of a camera department (dolly, truck, crane, rack focus) instead of describing what the shot should feel like.
CFG Scale
A dial controlling how literally a diffusion model obeys the prompt, trading obedience against image quality.
Character Consistency
Keeping the same face, body and wardrobe recognisably identical across shots, scenes and sessions.
ControlNet
A conditioning layer that constrains a diffusion model with structural input (a pose skeleton, a depth map, an edge trace) so composition survives a change of style.
Cost Per Accepted Asset
The only generation metric that means anything: total spend divided by the number of assets that actually shipped.
Deepfake
Synthetic media that depicts a real, identifiable person doing or saying something they did not, whether or not the intent is malicious.
Denoising Steps
How many passes a diffusion model takes to walk an image from noise to a finished frame.
Diffusion Model
The architecture behind most image and video generators: it learns to reverse a noising process, and generates by denoising random static into a frame that matches your conditioning.
Disclosure
Telling the audience that what they are looking at was made or materially altered by a generative model, in the place they will actually see it.
Drift
The slow, cumulative change in a subject across a set of generations (a face, a product, a colour) where no single frame looks wrong but the run does.
Fine-Tuning
Continuing a model’s training on a narrow dataset so it reliably produces one style, one subject or one house voice.
First–Last Frame
Handing a video model a start image and an end image and asking it to generate the motion between them.
Foundation Model
A large general-purpose model trained on broad data, intended to be adapted to specific tasks rather than used raw.
Identity Lock
Our pipeline for synthetic presenters: train the face once from a stills sheet, then run every subsequent variant from the trained identity rather than from a description.
Image-to-Video
Generating a moving shot from a still you have already approved, rather than from text alone.
Inpainting
Regenerating a masked region of an existing image while the rest of the frame stays untouched.
Master Plate
The single locked reference frame of a subject that every later generation is built from, so the subject cannot wander across a set.
MCP
An open protocol for connecting a model to external tools and data through a standard interface, so a capability written once works across clients.
Multimodal Model
A model that takes and produces more than one kind of input (text, image, audio, video) inside a single system.
Phantom Set
Our pipeline for any subject a customer can hold up against the picture: lock a master plate, specify the set once, then generate every angle from the plate.
Prompt Engineering
Writing model instructions with the precision of a brief: what, in what order, with what constraints and what it must not do.
RAG
Fetching relevant source material at run time and giving it to the model, so the answer comes from documents rather than from memory.
Reference Image
A picture supplied alongside the prompt to steer composition, style, colour or subject.
Register
A deliberate visual dialect: a decision about lighting model, edge quality and how much the world is allowed to be wrong.
Seed
The number that fixes a model’s starting randomness, making a generation reproducible.
Shot List
The ordered list of shots a piece needs, with framing, movement and intent stated per shot, written before anything is generated.
Style Transfer
Applying the visual character of one image or corpus to the content of another.
Synthetic Media
Any image, video, audio or text produced or materially altered by a generative model.
Synthetic UGC
Creator-style content (presenter to camera, phone-look, room lighting) produced with a trained identity rather than a booked creator.
Temporal Coherence
Whether a generated clip holds together across time: the same face, the same coat, the same number of fingers, from first frame to last.
Text-to-Video
Generating a moving shot directly from a written description, with no source image.
Token
The unit a language model reads and writes, roughly a short word or a fragment of one. It is also the unit most models are billed by.
Trained Identity
A face trained once from a sheet of stills, which then carries across sessions, models and formats without re-uploading a reference.
A term we should have defined and didn't?
Send it over. If it belongs in the working vocabulary of anyone shipping synthetic media, it belongs on this page.