Plan aspect ratio, length, audio before AI product videos

The guide is aimed at the moment before generation starts, not at prompt craft. Its five items are ordered as decisions that constrain each other: the placement fixes the aspect ratio, which fixes the composition; the length and the 2–3 seconds in which the product must be recognizable fix what each cut has to show; that shot list in turn makes the prompt easier to write.

The reference-image point speaks to a specific worry — damaging how the product looks. The guide says passing an actual product photo as a reference image is more stable than describing the item in text alone, and that you should also decide in advance whether the background changes or the product itself moves.

Audio is treated as a model-selection question rather than a finishing step: some recent video models can generate with sound, so whether you need only background music or also narration and effects changes which model you pick and what work follows. The comparison point closes the loop — swapping models inside one account is offered as the practical way to hold the prompt, reference image and length constant.