Dancing Cactus vs Baby Viral AI Video Prompt
A toddler in a cozy family living room pokes a dancing cactus toy, then looks shocked as it repeats the baby’s sounds. The moment turns into curious giggles and playful confusion.
Master prompt · 1,768 words
Create an **exactly 15-second vertical 9:16 ultra-photorealistic raw real-world family phone video** for **Seedance 2.5**, centered on a fictional **approximately 1-year-old baby from [TIER_1_COUNTRY]** having a funny spontaneous interaction with a real commercially available-style **green dancing talking cactus toy** that records and repeats nearby voices. The entire video must feel like a genuine parent unexpectedly captured a funny moment at home on an **iPhone 17 Pro rear/back camera**, never like an advertisement, movie, polished commercial, CGI animation, staged influencer video, or AI-generated scene. **COUNTRY / BABY VARIATION ENGINE:** For every generation, naturally rotate the household and baby appearance according to one Tier-1 market such as **USA, UK, Canada, Australia, New...
This prompt is written for Seedance 2.5. If you use a different model, keep the timing, phone-camera realism, and toy behavior, but simplify any very specific camera or sensor language if that model tends to ignore dense technical detail.
It assumes one continuous 15-second clip in vertical 9:16, starting already in the action. The frame is a parent-held medium-wide shot from about waist to knees up on the baby, with the whole cactus and pot always visible.
It is tuned for ultra-photoreal family phone footage, specifically messy, ordinary home realism rather than polished viral content. The model will most likely get the cactus wrong by making it too animated, too expressive, or moving off its base, and may also age the baby up or make the room look staged.
The lines you will most realistically edit are the country and city placeholders, plus the baby interaction choices: BABY_ACTION, BABY_SOUND_1, and BABY_SOUND_2. Those edits do most of the personalization without breaking the structure.
It assumes one continuous 15-second clip in vertical 9:16, starting already in the action. The frame is a parent-held medium-wide shot from about waist to knees up on the baby, with the whole cactus and pot always visible.
It is tuned for ultra-photoreal family phone footage, specifically messy, ordinary home realism rather than polished viral content. The model will most likely get the cactus wrong by making it too animated, too expressive, or moving off its base, and may also age the baby up or make the room look staged.
The lines you will most realistically edit are the country and city placeholders, plus the baby interaction choices: BABY_ACTION, BABY_SOUND_1, and BABY_SOUND_2. Those edits do most of the personalization without breaking the structure.
Questions about this prompt
Which models does this prompt work with?
It is written for Seedance 2.5. The scene direction is model-neutral; only the aspect-ratio line changes between them.
Can I use the output commercially?
Yes. Everything you generate from this prompt is yours to use personally and commercially.
What do I get when I unlock it?
The complete 1,768-word master prompt, copyable in one tap. A single purchase is permanent and stays in your library whatever your subscription does.
