What Is AI Casting (Synthetic Performers)?

AI casting refers to using artificial intelligence to select, simulate or generate performers, from hyper-real digital humans to fully synthetic video characters, for screen content and advertising.^[1] It spans talent search tools that rank real actors, avatar platforms such as Synthesia and HeyGen, and generative video systems like Sora or Runway that create fictional people on demand.^[2][3] For brands, AI casting raises questions about performance quality, consent, likeness rights and disclosure, which are now addressed in union agreements and advertising regulation.^[1][4]

Defining AI casting and synthetic performers

In advertising, AI casting covers two broad activities. First, algorithmic selection of human performers, where tools analyse showreels, demographics and past performance data to shortlist talent for a role.^[5] Second, the use of synthetic performers, where AI systems create or animate digital humans that deliver scripted lines, gestures and expressions in place of filmed actors.^[2] Synthetic performers can be linked to a real person’s likeness or be completely fictional. They are part of a wider field of synthetic media that uses generative models to produce photo-real or stylised people on screen.

Synthetic performers are distinct from traditional CGI characters because they are generated, animated or personalised largely through machine learning models, often from text prompts rather than frame-by-frame keyframing.^[2] This includes talking-head avatars with realistic lip sync, AI-driven facial animation mapped to a recorded voice, or fully generated video where every pixel is synthesised.^[3] For casting, the creative decision is less about which human to hire and more about which model, avatar style or digital likeness to deploy and how it should be controlled.

Types of AI performers: avatars and generative video

Hyper-real avatar platforms such as Synthesia, HeyGen, Soul Machines and Hour One create synthetic presenters that read scripts in multiple languages, with controllable voices, gestures and backgrounds.^[2][6] Brands use these avatars for explainer videos, customer support content and localisation where consistency and volume matter more than bespoke performance.^[2] Some avatars are built from an individual’s likeness under a licensing agreement, others are fully synthetic faces assembled from model training data. This raises questions about ongoing consent, re-use and where performance stops being a one-off service and becomes an enduring digital asset.

Generative video models like OpenAI’s Sora and Runway’s Gen-2 create moving people directly from written prompts or reference frames, without relying on a pre-defined avatar catalogue.^[3] These systems can depict fictional actors in any setting and can alter apparent age, clothing and behaviour frame by frame.^[3] In advertising, they intersect with AI recasting and face-swap localisation by making it easier to modify who appears in a scene after the shoot. Performance becomes partially programmable, which intensifies debates on authenticity and the boundary between simulation and acting.^[4]

Rights, unions and advertising regulation

Performer unions have begun to codify how AI casting and synthetic performers should be used. Equity, the UK union for performers, argues that AI systems cannot replicate “the human essence of performance” and calls for explicit consent, fair pay and contractual limits whenever a performer’s voice or likeness is captured for synthetic use.^[1] Its guidance highlights risks around perpetual use, synthetic replicas created from small data samples, and the difficulty of tracking how digital doubles are reused across campaigns.^[1]

In the United States, SAG-AFTRA’s 2023 TV/Theatrical agreement established the “Three Cs” framework for AI use: Consent, Compensation and Control.^[4] Consent requires clear permission before creating a digital replica; compensation requires payment that reflects the economic value of any synthetic replacement; and control covers how, where and how long a replica can be used.^[4] These principles are increasingly referenced in global brand and agency contracts, even outside US union productions, as a baseline for ethical AI casting.

Advertising regulators are also responding. The UK Advertising Standards Authority (ASA) has stated that AI-generated content is subject to the same rules on misleadingness, harm and offence as any other ad, and that manipulative or synthetic techniques must not materially mislead consumers. While there is not yet a standalone AI casting code, emerging guidance stresses transparency when significant elements of an ad, including presenters, are artificially generated. For brands, that means assessing whether synthetic performers could give a false impression of product experience, endorsement or typical results.

Brand use cases and practical considerations

Brands use AI casting to create multilingual spokesperson videos at scale, to personalise dynamic creatives, and to prototype scripts before committing to a live shoot.^[2][6] Synthetic presenters can reduce production time for how-to or support content and make it easier to update messaging without re-booking talent.^[6] Some advertisers also test narrative ideas with AI-generated performers before engaging human casts, treating synthetic work as a pre-visualisation stage rather than final output.^[5]

However, there are reputational and creative trade-offs. Analysts at the World Economic Forum argue that labelling synthetic figures as “actors” risks flattening storytelling, eroding the value of human work and confusing audiences about what is authentic.^[5] Voice-actor bodies such as NAVA similarly warn that unchecked vocal cloning can undermine professional performers unless consent and compensation are robustly enforced.^[6] For long-term brand building, many marketers are therefore experimenting with AI casting in tightly bounded contexts while continuing to rely on human performers for emotionally rich, high-stakes campaigns.^[5]