Reference to Video is an AI generation tool that creates new video clips from a reference video and user provided photos. The reference video supplies the movement, camera motion, and timing, while the photos introduce the subject and visual details for the output. This method enables the recreation of specific actions or camera sequences without relying solely on text prompts to describe complex motion. The result is an AI interpretation that requires review of the subject, framing, and transitions. The workflow begins with a template from the library or an uploaded custom clip. Users add clear photos of the subject and can use a custom prompt to describe desired changes to the setting, subject, or camera direction. The tool includes multiple generation models, such as Seedance 2.0, Fast, and Mini, with adjustable settings for resolution, length, and audio. Inputs, settings, and results are managed in a single workspace. Generation progress can be tracked in the My creations section, and the final version can be downloaded. Typical use cases include recreating fashion shots, product reveals, camera moves, and expressive performances with different subjects or wardrobes. The platform hosts community trend pages for comparing original clips with AI recreations. Best practices include using a reference video with a clear beginning and ending action, and ensuring full body photos include the hands and feet. Results can be refined by changing one input at a time. The generation process requires acceptance of the platform use policy and confirmation of media permissions.