The visuals and voice of the character will be fully generated using generative AI techniques. The dialog can be written by a person, but if so, it must start from text. That is, it can't take voice captures and stylistically transfer them (otherwise an actor would be imparting emotion and other information.)
Note: live-action rules out cartoons and anime.
1,000
3.00@jonsimon good question. See the clarification above. The intent was to make sure that the acting was really coming from the actor. An actor adds a lot of emotion and visceral feel between text and the recording. If all that's done is stylistic transfer on the voice, then it's just a transform. The point here is that the AI is doing the acting.
@RealityQuotient Thank you for the clarification. Can there be post-hoc editing/touchups of the AI-generates audio or visuals? Many of these methods leave telltale artifacts.
@jonsimon Too broad to answer. I also suspect it will be hard for us to understand the full scope, unless there's a "making of". But I'll give some examples.
Meryl Streep famously gives take after take after take, each one very different, allowing the director to decide. That sort of curation is fine, it's what directors do. And, directors do give hints, "e.g. this time show some anger". But if it's more like modern CGI, or stop motion, where none of the artistic flare comes in, then it won't count for this.
The real intent behind this is, will actors become fungible. If characters can be created, and reused, then TV and movie content will become cheaper since the studios won't have to pay up for successes as much, they'll own the IP. You see this with cartoons to some extent. The question is whether it can map to live-action.
@RealityQuotient Ahh ok I understand the intent much better now. So then, is it alright if some graphics designer has a heavy hand in the crafting of the image and voice, so long as there's some AI involved in the final product? The most plausible path I see for this is something like an initial 3D model that gets made by a human, which is then made l ultra-realistic by an AI. And whose voice is based on a few samples from same real-life person(s) which can then be extended to say whatever is desired using AI technology. Would that count, or too much human intervention?