Ozil Goal AI Prompt for Kling and Seedance (Copy & Paste)

These two prompts are a do-it-yourself route, tested on 17 October 2026 with a person and a dog. Use them in any image-to-video model that accepts a start frame (Kling, Seedance, Grok Imagine and others). Results vary between models and takes, so plan on a few generations.

Step 1: make the first frame

Generate an empty floodlit stadium first (9:16), then use an image model that accepts two reference images: the stadium as the first image and your photo as the second.

Stadium prompt (text-to-image, 9:16):

Photorealistic vertical phone-video still at a big modern football stadium at night, bright white floodlights, packed stands blurred in the background, wet green grass. Camera low and behind the attacker's position, about 12 metres from goal, looking toward the goal. A goalkeeper in a plain neon-green kit stands in front of a plain white goal. Two defenders in plain navy-blue kits with no badges or logos run in from the left and right edges. The lower middle third of the frame is empty grass with a white football on it. No team crests, no sponsor logos, no advertising text.

First-frame prompt (image edit, stadium + your photo):

Use the first image as the exact scene: keep the night stadium, floodlights, goal, goalkeeper, both defenders, camera angle and lighting unchanged. Take the main subject from the second image (a person, a pet or a character) and place them in the lower middle of the frame as the attacking football player, with the white ball right at their feet or paws, body angled three-quarters so their face is clearly visible to the camera, taking about one third of the frame height. Dress a person in a plain white football shirt and white shorts with no badge, logo or number; an animal keeps its natural look. Keep the subject's face, hair and fur exactly as in the second image and match the cold white floodlight lighting on them. Photorealistic unless the subject is a cartoon, no text.

Step 2: the timed video prompt

Use the first frame as the start image, 9:16, 15 seconds if your model allows it. Shorter clips can’t fit all four beats; with 10 seconds, drop the celebration.

Handheld vertical phone video of a floodlit night football match, one continuous shot, the camera tracks the player at pitch level. 0-3s: the player turns toward the goal and sprints, dribbling the white ball close. 3-6s: the goalkeeper in green rushes out; the player calmly chips the ball over the goalkeeper, it bounces once in the box. 6-10s: chasing the ball, the first defender in navy slides in hard from the left and misses, then the second defender slides in from the right and misses too, both skidding past on the wet grass. 10-13s: the player stops the ball on the goal line, waits a beat, and calmly taps it into the empty net. 13-15s: they turn to the camera and celebrate, the crowd roars. If the player is an animal it does the same beats with its paws and nose and celebrates by jumping and wagging. Realistic ball physics, crowd noise, keep the player's face, fur and outfit consistent. Pitch-side advertising boards are plain and blank: no brand names, logos, crests or text anywhere.

Kling vs Seedance vs Grok

• Kling: strong motion and faces; 10-second clips on many plans, so trim the timeline.
• Seedance: good camera tracking; check the clip length your plan allows.
• Grok Imagine: 15-second 9:16 clips; this is what we tested the prompt with.

Fixes for common problems

Problem	Fix
Ball disappears in the first frame	Add "the white ball right at their feet" and regenerate the frame only
Face drifts during the slides	Use a sharper, front-facing photo; keep "face clearly visible" in the frame prompt
Real brand logos on ad boards	Keep the "plain and blank advertising boards" line
Player shoots instead of chipping	Move "chips the ball over the goalkeeper" earlier and keep the timestamps
Club colours appear	Say "plain white shirt, no badge" and avoid naming any club

Don’t name real clubs, players or competitions in the prompt: models may refuse, and real crests in your video can get it taken down.

Skip the prompting

Prompts alone struggle with a four-beat move like this one. Ozil AI uses a different method: a reference video of the goal (our own re-enactment) plus your photo, so the run, the chip, the two missed slide tackles and the finish follow the real timing. One photo, no prompting, from about $2.50 per video.