The host's claim is that Nano Banana Pro understands prompts well enough that a compact or non-verbal input can replace a long written description. The source shows four such inputs:
The three demo scenarios are a studio portrait moved into new environments, a fashion sketch turned into a photo, and a lighting sketch turned into a cinematic scene.
Before writing a paragraph of description, ask whether a reference image, a sketch, a diagram or a coordinate pair carries the same information. Add it as a reference (see Nano Banana Pro in Photoshop via the Luminel plug-in for why extra reference images need the plug-in) and keep the text prompt to the action and the constraints.
Compare minimal-video-prompt-references-hold-identity, which makes the same move for video, where references hold identity so the prompt can stay short.