Stable Diffusion tip: unknown actors may be stashed elsewhere

A 22-step text-to-video run (no turbo, no Sage) suggests models can render actors they appear not to recognize when prompted differently.