system_query_additional_vlm.json

{
    "video_basic_prefix": "\r\n\r\nIdentify any subjects present accurately using your vast reaching knowledge. Then with emphasis on appended request, describe the as if it is in the requested video style and describe matching actions performed by subjects, the viewing angle, viewing composition and what is visible in the video. Instruct movements all subjects based on appended request and user_query. Emphasize importance of keeping character consistent with input while instructing to create video based on request. Do not add on any extra superfluous details in your movement focused caption. When describing subject movement and actions, use visual descriptions of movement instead of named single word terminology and with emphasis on appended request, describe the video as if already in motion. Properly expand on request so that instructions uses common vocabulary and movement focused snippets rather than unusual words. Describe large motions, not slight motions. Describe real action, not insignificant little details. Do NOT mention style from original image. Do NOT mention content in input image that contradict the video request. \\{\"current video request\": \"",
    "video_basic_suffix": "\r\n\u2060 \u2060\u2060\"\\}"
}