# Zombie AI Prompts: Transformation, Mood and Face Consistency

Write a focused zombie video prompt, prepare a suitable reference photo and diagnose common scene failures.

Select the matching video type in the studio, verify your email and use credits to generate. Review the entire result before sharing.

## Start with one visible change

The studio offers seven preset video types, without a custom prompt field. Copy these texts into a compatible custom-prompt tool.

Decide whether the clip is about changing appearance or telling a story. Trying to transform a face, run through a crowd and complete a reunion in the same short shot gives several different instructions equal importance. Begin with the smallest scene that conveys your idea.

## Prepare the reference

Choose a photo where the face is sharp, visible and large enough to read. Leave space around the shoulders for a slow turn. Glasses, hair across the eyes, extreme side views and strong shadows introduce details you will need to check carefully. These are practical preparation suggestions, not findings from a controlled model test.

Crop for the intended frame before generating. A portrait composition and a wide street scene have different needs. Do not rely on a resolution label to repair a face that is too small in the input. Use an adult subject and a photo you have permission to upload.

## Build and test the prompt

Name the person through the reference, describe the setting, specify the appearance change, choose one action, then set camera and lighting. Describe visible effects such as pale makeup instead of vague quality slogans. The prompt below keeps the transformation non-graphic and the camera stationary.

Use the text as a starting point. Replace the street, jacket or expression to fit the actual photo. Keep the subject and movement instructions unchanged for your first comparison; otherwise you will not know which edit affected the result.

A continuous shot of the adult person in the reference photo in a quiet street at dusk. Keep recognizable facial features and the same jacket. Their skin slowly becomes pale with subtle theatrical zombie makeup. They turn toward the camera once, then hold a hesitant smile. Static camera, cool light, light mist. Natural anatomy. No blood, injuries, weapons, text or logos.

## If the face or movement changes

Watch the whole clip once, then pause where the face first stops resembling the reference. Separate an appearance change you requested from unwanted identity drift. If the turn causes the failure, compare a smaller head movement. If hands deform, remove reaching or waving and try a quieter pose.

Change one instruction per comparison and retain the original task ID and settings. A new generation can vary even with the same text. Do not resubmit a paid request merely because the page timed out; first check whether the original task exists. No prompt can guarantee an identical face in every frame.

## Finish the clip

Review the opening and ending for abrupt jumps, then check hands, facial details and background continuity. Add a soundtrack only after the visual timing works. Use music and images you have permission to publish, and describe a realistic AI scene as synthetic.

A useful first attempt is one understandable transformation. Once that works in the tool you choose, develop another shot rather than continually adding instructions to the original.

## Questions before your first scene

### Can I paste this into the studio here?

The studio offers seven preset video types, without a custom prompt field. Copy these texts into a compatible custom-prompt tool.

### Should I request several actions at once?

Start with one action, then compare a separate shot for the next beat. This makes revisions easier to interpret.

### Can negative instructions guarantee a clean result?

No. They express the intended scene; you still need to inspect the delivered clip.