A face thatholds the wholeperformance.

A talking head needs a mouth to match words for thirty seconds. A performance needs a face to carry a whole track: phrasing, emphasis, the breath before a line, the look between bars. It also has to stay the same face across every cut, which is where most generated video gives itself away. Someone who looks subtly different in the third verse reads as a different person, and the audience stops following the argument and starts watching the artefacts.

On the work shown here, the client does. Radical Discourse writes the songs and produces the series. Our part is the AI performance and the shot work. On other projects we write too, but it is worth separating, because the writing is what makes something worth watching and claiming someone else's is not a service we offer.

Anything where a performance has to be delivered by someone who cannot be in the room: a musical performance, a historical figure, a brand character, a founder who is not available for a shoot, or the same piece delivered in several languages by the same face. The constraint is not the genre. It is whether the thing is worth performing in the first place.

With written consent from the person, yes, and only with it. Likeness and voice work here happens on paper first. We track the legal position across Texas and elsewhere so a legal team gets a clear answer rather than a surprise, and if the consent is not there, neither is the work.

Fifty-plus, with the performance carried natively rather than dubbed over the top, so the mouth matches the language being spoken. For an organisation publishing the same thing across markets, that is usually the part that makes the economics work.

Scoped per project, and the honest drivers are length, how many characters have to stay consistent, and whether the material already exists. We will quote after a conversation about the actual job rather than before it.