提示词

Crazy Frog AI 提示词

笑点在于骑手身下空无一物。提示词应明确哪些物件不可见、区分两个角色，并给出简短且连续的动作。

Create a 15-second, 9:16 comedic invisible-motorcycle chase.

@image1 is the only reference and supplies the rider identity. Preserve their face, hair and outfit. Create a fictional helmeted adult officer with a distinct identity.

The rider rides empty air: no visible bike, wheels, saddle or handlebars. Only the officer has a visible futuristic hover bike.

Use original energetic instrumental audio, wind and engine-like effects. No recognizable songs, copied melodies, lyrics, dialogue, logos or watermark.

3s: Tracking full-body shot on a sunlit coastal overpass, turquoise sea and warm sunset. The rider crouches in mid-air, knees bent, fists gripping invisible handlebars. Absolutely no bike, wheels, seat or handlebars beneath the rider. Keep the rider recognizable.

2s: The rider glances back with a cheeky smile. A fictional helmeted officer approaches in the distance on a small futuristic hover bike. Keep the rider floating in the same riding pose; no visible motorcycle beneath the rider.

3s: Low side tracking shot. Rider and chasing officer speed along the same road. Wind moves their clothes; road markings streak past. Keep both identities distinct, natural anatomy, no collision or weapons.

2s: The rider performs one playful, physically impossible backward flip in the air. There is still no vehicle beneath the rider. The officer follows on the hover bike at a safe distance.

2s: The rider lands back in the original invisible riding pose, briefly throws both hands up in celebration, then reaches forward to invisible handlebars. Preserve face, clothing and road continuity.

3s: Rear three-quarter tracking shot. The rider accelerates away on the invisible motorcycle toward the horizon. The hover-bike officer falls behind. End on a clean wide shot, no crash, text or watermark.

让摩托真正隐形

写明“骑手身下没有可见车轮、车座或车把”。追逐者可以骑可见的悬浮摩托，所以隐形要求要明确指向骑手。

保持骑手可辨认

用一张清晰骑手参考照，要求保留面部、发型和服装。追逐者由 AI 创作为戴头盔的虚构人物，全程保持两个人物身份独立。

把动作说清楚

从隐形骑行开始，展示追逐，加入搞怪翻转，再以加速离开收尾。时间段是给模型的指引，实际动作和相似度可能变化。