SocialToPrompt SocialToPrompt

The "Last Wave" Kiss Is Everywhere: Steal Its Exact AI Video Prompt (Plus 4 More Trends to Copy This Week)

Author: SocialToPrompt Team Date: 2026-11-03 09:44:26
The "Last Wave" Kiss Is Everywhere: Steal Its Exact AI Video Prompt (Plus 4 More Trends to Copy This Week)

Published November 3, 2026

Open TikTok right now and you will see it: a couple, silhouetted against a sinking skyline, leaning in for a slow-motion kiss while a wall of water — or a meteor, or an explosion — rushes up from the horizon and swallows the frame. Creators are calling it the “Last Wave”. Some call it “the last kiss against disaster.” It has racked up hundreds of millions of views across Reels, Shorts and Douyin in the last two weeks alone, and it is not going anywhere.

You have also probably seen the hyper-realistic “Nano Banana” selfies, the Italian brainrot shark in Nikes, and those impossibly smooth Veo 3 drone shots through misty forests that nobody can tell are AI.

Here is the thing nobody tells you: none of these trends are won by the people who write the best prompts from scratch. They are won by the people who reverse-engineer the videos that are already going viral and feed them into the next model.

Let’s break down the five formats dominating right now, the prompt anatomy behind each, and — most importantly — how to skip the guessing and pull the exact working prompt out of any viral video in about thirty seconds.

1. The “Last Wave” — slow-mo kiss against a tsunami

This is the trend of the moment. A couple ignores an oncoming catastrophe — a colossal tsunami, a meteor strike, a mushroom cloud — steps closer, embraces, and kisses in slow motion right as the wave engulfs the screen. The emotional hook is fatalism plus intimacy; the visual hook is the extreme scale contrast between the tiny human moment and the apocalyptic background.

Why it works: It is a universal, instantly readable story beat. No dialogue. No text needed. It translates across every language on earth, which is why it is blowing up simultaneously in English, Russian, Spanish, Mandarin and Indonesian feeds.

Prompt anatomy (the part people hand-write badly):

In the background, a colossal tsunami wave rises and destroys the city.
The camera slightly shakes from the shockwave. A couple ignores the danger,
steps closer, hugs, and shares a passionate, slow-motion last kiss right as
the wave engulfs the screen. Epic cinematic motion, photorealistic, 35mm,
shallow depth of field, warm golden-hour backlight on their faces.

The details that make or break it: camera shake from the shockwave, the ignoring of the danger (this is the emotional beat), and the timing of the wave hitting the screen exactly on the kiss. Miss any of those and you get a generic disaster clip.

2. “Nano Banana” hyper-real selfies

Late August 2026, a prompt template called “Nano Banana” started circulating on X. It instructs image models to replicate a specific person’s face from a single reference photo while mimicking iPhone 16 Pro Max camera artifacts — skin grain, lens flare, slight exposure bias — specifically to defeat “is this AI?” detectors. By November it has been remixed into video: people are generating moving selfies that look like raw camera roll footage.

We are not going to provide the identity-replication portion of the prompt here — that is a different conversation with real deepfake risks. But the style part is fair game and worth copying:

Prompt anatomy:

Ultra-realistic front-camera selfie, iPhone 16 Pro Max front lens,
slightly overexposed indoor lighting, natural skin texture with visible pores,
subtle motion blur from handheld stabilization, no beautification, no filter,
candid expression, 24mm equivalent.

That last line — no beautification, no filter — is the secret. Every AI selfie defaults to influencer skin; the ones that look real explicitly forbid it.

3. Italian brainrot — the biggest style of 2026

Tralalero Tralala — the shark wearing three blue Nikes — crossed several billion cumulative views this year. The format is absurd on purpose: improbable animal hybrids wearing incongruous human accessories, with pseudo-Italian synthesized voiceovers (“brrrt, bambino, pallino…”). It looks dumb and it is intentionally dumb, which is exactly why it loops.

Prompt anatomy:

Photorealistic 3D render of [improbable animal hybrid] wearing [incongruous
human accessory], standing in [bizarre location], dramatic side lighting,
hyper-detailed fur and fabric textures, slightly stiff character pose,
looking directly at camera. Add a synthesized Italian-male voiceover in
broken English mumbling a made-up food word.

The winning variable here is consistency. The accounts that scale lock one character design and post 4-7 clips a day with the same character in new scenarios. One-off clips barely move.

4. Veo 3 “you can’t tell it’s AI” cinematic B-roll

With Sora wound down in 2026 (app retired in April, API shut in September), Veo 3 and Kling 2.x have absorbed the creators who want photorealistic, near-invisible AI footage. The format everyone is chasing: a drone gliding through a misty forest, a slow push-in on espresso pouring into a ceramic cup, a rainy Tokyo crosswalk at 2am. No plot. No characters. Pure mood.

Prompt anatomy:

[Cinematic camera move] through [location], [time of day] lighting,
[film stock or director reference], shallow depth of field, photorealistic,
no people, ambient motion only — leaves drifting, steam rising, rain streaks.

These clips get embedded inside YouTube videos, TikTok edits, and ads as B-roll. They also rack up millions of standalone late-night views because they are perfect autoplay-loop background material.

5. AI lo-fi loops

Close cousin to #4, but deliberately cozy and loopable: a girl reading by a rainy window, a cat watching snow fall, a coffee shop at 3 a.m. No narrative arc, just slow ambient motion designed to play on repeat forever. Massive on YouTube Shorts and Instagram Reels.

Prompt anatomy:

Cozy [scene], slow ambient motion only — rain streaks on glass, steam from
a mug, flickering candlelight, warm interior color grade, no camera movement,
loopable on itself, lo-fi film grain.

The real bottleneck isn’t the model — it’s the prompt

Here is what the OpusClip trend roundup, every TikTok AI creator, and every brand social team has quietly converged on in 2026:

  • The tools are no longer the constraint. Veo 3, Kling, Seedance 2.5, Runway Gen-4, Luma, Pika, Hailuo, Vidu and the rest are all within a usable range of each other.
  • The constraint is translation: watching a 12-second clip that makes you cry and being completely unable to articulate why it works well enough to rebuild it.

Hand-writing prompts for these trends is a skill that takes months to build. You have to learn to notice the camera move, the lens choice, the lighting direction, the timing of the cut, the color grade — while you are watching a video that is deliberately designed to feel fast and invisible.

Stop writing prompts. Steal them.

This is the whole idea behind SocialToPrompt. You already have the raw material: the video that just popped up on your FYP. You do not need to stare at it and guess.

How it works:

  1. Paste the link. Drop any video URL from YouTube, TikTok, Instagram, X, Vimeo, Bilibili, Xiaohongshu, Reddit, LinkedIn, or any of the 30+ platforms supported. You can also upload a local MP4/MOV/WebM file.
  2. The AI watches every frame. It extracts the motion, camera movement, composition, lighting, color grade, timing and aesthetic — the same variables the prompt anatomy lists above.
  3. Get back a structured prompt. A detailed, ready-to-paste prompt you can drop straight into Veo, Kling, Seedance, Runway, Luma, Pika, Hailuo, Vidu, Veonib, or any other text-to-video tool — and then remix it: change the couple, swap the tsunami for a meteor, change the city to Tokyo, run it with your own face.

That last part is the multiplier. The “Last Wave” template works with a tsunami, an explosion, an alien invasion, a sandstorm, a solar eclipse. Once you have the structured prompt, you have the engine — not just the clip.

Which model should you paste the prompt into?

The prompt you extract from a viral video is model-agnostic text, but it helps to know where it lands best:

  • Veo 3 — for the photoreal cinematic B-roll and the Last Wave realism.
  • Kling 2.x / Seedance 2.5 — for dramatic camera motion, slow-motion, and the longer 15-30 second clips.
  • Runway Gen-4 / Pika — for stylized and anime-leaning looks.
  • Hailuo / Luma / Vidu — for character consistency and character POV clips.
  • Veo, Gemini Omni, Wan, Hunyuan, SkyReels, Mochi, CogVideoX, StepVideo — all accept the same structured prompt format.

Pick one trend. Ship it this week.

The single most expensive mistake creators make in 2026 is trying to chase all five of these trends at once. Pick the one that fits your account, find the top 5 videos doing it right this week, paste each link into SocialToPrompt, and rebuild them with your own twist. Ship daily for two weeks. The analytics will tell you in fourteen days whether your audience wants it — not fourteen months.

Ready to steal your first prompt? Paste a viral video link into SocialToPrompt and get a structured, model-ready prompt in seconds. Every new account starts with 3 free credits — no credit card required.


What viral trend should we reverse-engineer next? Drop a video link in the comments and we’ll turn it into a prompt.

Share Article

Related Articles

Recommended Reading

Ready to Get Started?

Experience our product immediately and explore more possibilities.