One portrait + audio → a realistic talking presenter with OmniHuman.
Press generate, output appears here