What Is an AI TVC?

An AI TVC is a television commercial in which some or all visual content is created using generative artificial intelligence, typically as part of a wider TVC production workflow. Between 2024 and 2025, brands and agencies began testing tools such as Runway Gen-3 and Gen-3.5, Google Veo, OpenAI Sora and Chinese model Kling for broadcast-quality, photoreal video, often alongside conventional live action and CG pipelines.^[1][2] This raises new questions around rights, disclosure and compliance with Clearcast and ASA rules in the UK.

Definition and scope of an AI TVC

In practice, an AI TVC is any television commercial intended for linear or BVOD broadcast where generative models create a material part of the picture, whether entire scenes or specific elements such as environments, crowd extensions or product shots.^[1][2] The audio may remain traditionally produced, for example recorded VO and music, or may also use synthetic voices. Unlike purely experimental AI clips, an AI TVC must conform to broadcast technical specs and advertising rules, so it is produced within established spot-delivery workflows and cleared by bodies such as Clearcast for UK TV placement.^[3]

AI TVCs sit on a spectrum. At one end are fully AI-generated films where every frame comes from a model such as Runway Gen-3, stitched and graded in post. At the other are hybrid executions where AI augments live action plates, similar to VFX or animation. Agencies increasingly use AI for previsualisation and AI animatics first, then selectively bring those tools into final production when brand, legal and technical risks are understood.^[2][4]

Key generative video models in 2024–2025 workflows

Runway’s Gen-3 and follow-up Gen-3.5 models are among the most widely used in commercial contexts, positioned by the company as foundation models for expressive human characters, camera motion and multi-shot sequences suitable for professional production.^[1] They support text-to-video, image-to-video and video-to-video workflows, which makes them adaptable to both fully synthetic spots and hybrid pipelines that extend or re-version live action. Coverr’s production-focused deployment of Runway Gen-3 similarly emphasises temporally stable, artefact-light clips for short-form advertising assets.^[5]

OpenAI’s Sora, announced in 2024 as a text-to-video model capable of generating high-resolution, photoreal sequences with consistent characters and complex motion, has been tested by selected brands and directors for conceptual TVC work, although full broadcast deployment remains limited while access is restricted and legal frameworks evolve.^[2] Google’s Veo, previewed within the Vertex AI platform, targets cinematic video with camera control and higher-fidelity motion,^[6] while Chinese model Kling has been showcased for long, photoreal shots that resemble live action.^[4] These models are particularly relevant where advertisers seek photoreal AI video rather than stylised animation.

In most UK and EU TVC workflows, these systems currently augment rather than replace traditional production. Agencies may generate establishing shots, product worlds or abstract visual metaphors in Runway or Sora, then composite them with live action footage in familiar tools such as Nuke or After Effects. This hybrid approach helps teams maintain control over performance, continuity and brand assets while benefiting from rapid exploration and lower-cost alternatives to some location or CG-heavy work.^[1][4]

Hybrid AI–live action pipelines and craft considerations

Hybrid pipelines generally follow a familiar production structure. Scripts and boards are developed as usual, often supported by AI animatics or test clips. Selected shots are earmarked for AI generation, with detailed prompts, reference frames and layout guides created by directors, VFX supervisors and prompt specialists. Generated passes are then brought into a standard post pipeline for clean-up, grading, type design and audio, preserving the broadcast QC and finishing practices used for non-AI TVCs.^[2][4]

Craft concerns focus on temporal consistency, character likeness and product accuracy. Runway Gen-3’s multi-shot and keyframe tools, for example, are used to keep wardrobe, hair and props coherent across a sequence.^[1] For regulated categories such as food, finance or healthcare, agencies report using AI mainly for backgrounds or atmospherics where factual claims are not at stake, keeping depictions of dosage, pack formats or legal information grounded in filmed or 3D-verified assets.^[3] This partitioning of AI and non-AI elements can simplify substantiation and rights management.

Clearcast clearance, regulation and early case studies

Clearcast assesses AI-generated commercials under existing BCAP and CAP rules, focusing on substantiation, misleadingness and harm rather than the tools used.^[3] Their guidance stresses that advertisers must hold evidence for all claims, ensure AI-created visuals do not exaggerate product performance, and secure rights for any training data or likenesses used.^[3] The ASA has separately signalled that AI-assisted ads remain fully subject to misleadingness and social responsibility provisions, pointing to concerns such as deepfakes, synthetic endorsements and body image manipulation.^[7]

By 2024, UK and European broadcasters had begun airing AI-supported commercials. Campaign reported on brands such as Nestlé’s Milkybar using AI for set and background generation while retaining filmed talent and products, as well as smaller advertisers running fully AI-originated spots on BVOD where media budgets are modest and creative needs to stretch across multiple edits.^[2] Trade press coverage of early Sora and Runway-based films for automotive, fashion and entertainment clients highlights both cost efficiencies and the need for increased legal review at script and animatic stage, particularly around likeness rights and synthetic crowd scenes.^[2][4]

For practitioners, the implications are straightforward. AI usage does not remove the requirement for Clearcast pre-clearance or technical certification, and stations will still reject spots that fail delivery specs regardless of how they are produced.^[3] Robust documentation of prompts, model settings and post-processing choices is increasingly being treated as part of the production bible, so legal, client and compliance teams can evidence how any contentious frames were produced if challenged by regulators or the public.^[3][7]