When we work on your first AI spot, words like keyframe, hook, avatar or lip-sync come up. Here they are, explained in a few sentences each, no padding. Ordered the way they appear in production: from the idea through the making to delivery and the law.
Basics
AI ad
An ad whose picture, video, voice or music was created with generative AI. Led by production people: direction, editing, colour and sound remain a craft. How it works with us in practice is in the article How an AI ad spot is made.
Generative AI
Models that create new images, video, sound or text from a text or image prompt. In advertising they replace part of the production, not the idea or the direction.
Prompt
The text instruction for an AI model: what is in the shot, which camera, light and mood. A good prompt is closer to a shot script than a single sentence. Hence our claim “Ads written in prompts”.
AI video and AI spot
An ad video generated by AI from keyframes and prompts, cut and scored like a regular spot. Usually 15 to 30 seconds. What it costs is on the AI video production page.
Text-to-video and image-to-video
Two modes of video generation: from a text prompt, or from a starting image. Advertising mostly uses the second, because it holds the product and the composition.
Preparation
Storyboard
A drawn sequence of shots with a description of the action. An AI storyboard is made from a brief within days and shows the film before production starts.
Animatic
A rough animated version of the storyboard with timing, music and sound. It checks rhythm and length before production.
Moodboard
A collage of references that sets the style: colours, light, type of shots. The AI brand kit grows out of it.
AI brand kit
A set of rules and references for the model: colours, typography, tone, typical scenes. It keeps the outputs on brand even months later.
Reference image
An image the model stays faithful to while generating: product shape, logo, the avatar's face. In practice a reference beats a verbal description, so it is chosen more carefully than the prompt.
Keyframe
A still image from which the model generates motion. Composition, product and light are approved on the keyframe before any video exists. A fix on an image is far cheaper than a fix in video.
People and voices
AI avatar
A synthetic face with a consistent identity across videos. It speaks several languages and the next video needs no new casting. Details on the AI avatars and UGC videos page.
UGC video
A video in the style of user content: a person talking, phone in hand, a natural setting. Made with creators, or with an AI avatar. The prices of both routes are compared in UGC videos: what they are and what they cost.
Lip-sync
Matching the movement of the mouth to the voice. In AI dubbing a dedicated model handles it so the speaker appears to speak the new language.
AI dubbing
Translation, a synthetic voice that keeps the character of the original, and lip-sync. The finished spot speaks another language without a new shoot. How it works and where it falls short is in AI video dubbing.
Localization
A language version of a spot for another market: dubbing or subtitles, on-screen text replaced, formats adjusted. Music and edit stay. The service: AI dubbing and localization.
Deepfake
A synthetic image or voice imitating a real person. In advertising only with that person's consent and with labelling. Without consent it breaches their rights; without a label it breaches the AI Act.
Cuts and formats
Hook (opening hook)
The first two to three seconds of a spot, which decide whether the viewer stays. That is why several hooks are tested on the same spot; with us three are included.
Cuts and lengths
Versions of the same spot: full length and a six-second cut, vertical, horizontal and square. Six cuts with us means three hooks times two lengths, not six different videos.
Performance package
A set of dozens of video variants (30 with us) for Meta and TikTok that are tested against each other in the campaign. The variant with the best results wins, not the prettiest one.
Upscale and mastering
Enlarging the output to a higher resolution and finalizing colour and sound. A spot is delivered in HD, the chosen final in 4K.
Artefacts and hallucinations
Model errors: extra fingers, broken text on packaging, an object that changes shape. That is why every shot is generated several times and only the one without a flaw goes into the edit.
Photos
Packshot
A product photo on a clean, usually white background. The basis for an online shop and the reference for AI-generated scenes.
Product visual (lifestyle scene)
The product in a setting: on a table, by the water, in snow, in a hand. From one packshot AI creates dozens of them. The service: AI product visuals.
Spec concept (spec spot)
A showcase spot for a known product without a brief from the brand. It shows the studio's craft, it is not a client commission. That is exactly what the showcase on our main page is.
The law
EU AI Act
The European regulation on artificial intelligence. From 2 August 2026 it requires synthetic content, including advertising, to be labelled. What it means in practice is in How to label AI content under the EU AI Act.
AI content labelling
A machine-readable mark in the output and, for realistic content, a visible notice for the viewer. An obligation under the EU AI Act that we handle for the client.
Read next
- How an AI ad spot is made: the process step by step and what AI cannot do yet.
- How much does an AI ad cost in 2026?: spots, visuals and content prices in one place.
- AI ads by well-known brands: what worked and what drew criticism.
Missing a term?
Tell us. And if you are planning your first AI spot, send us two sentences about what you sell. We reply within 24 hours.
Book a call →