Runway's Gen-1 research page describes video generation that starts with an existing video. An image or written prompt supplies the desired appearance, while the source video supplies structure. This lets a creator use recorded footage or a rough render as the starting material for a different visual treatment.
The February 2023 announcement also describes applying that treatment only to selected regions. The central capability is transforming an existing shot with text or image guidance, giving the creator source material to anchor the result.
