Use this prompt with an image-to-video model that supports reference images:
Create a cinematic, emotional zombie-apocalypse reunion featuring the two reference subjects. The first subject is a human survivor and the second is a recognizable, non-graphic zombie version of the other subject. In a ruined city, the survivor hesitates as they recognize their loved one. Instead of violence, they move toward each other and share a heartfelt embrace. Preserve facial identity as consistently as possible, with natural anatomy, believable expressions, gentle camera movement, atmospheric cinematic lighting, and an emotional ending. Avoid gore, graphic injury, fighting, distorted faces, extra limbs, duplicate subjects, and text overlays.
Keep identity and action instructions concrete. Specify which reference is the survivor and which is the zombie. If your model supports aspect ratio controls, use 9:16 for a vertical social clip or 16:9 for a landscape sequence. Do not request a duration that the selected model cannot produce.
For a longer love-story effect, generate two shots rather than forcing every beat into one request:
Edit the successful shots together after generation. This is usually more reliable than asking one model call to preserve identity while changing time, wardrobe, location, and mood.
When the model provides a separate negative-prompt field, use: gore, blood, graphic wounds, violence, deformed face, extra fingers, extra limbs, duplicate people, identity drift, sudden cuts, captions, logos, watermark.
No prompt guarantees a particular output. Review every generated video before publishing it, especially when it depicts a real person. Continue with the step-by-step tutorial, try the focused Zombie Hug AI template, or read the Zombie Love Story guide.