# AI video trends this week — what people are remaking, and how to make yours

> The viral AI video trends of the moment, with the original clip, why each one spreads and a ready template to make your own version with your character. Updated when a trend lands.

Published 2026-10-10 · eroq.ai — canonical: https://eroq.ai/blog/ai-video-trends-this-week


Updated 10 October 2026. This page is the running list of the AI video trends worth remaking right now: the clips the feeds are full of, each with its origin, what makes it work, and the eroq template that puts your own character in it. New entries land within a day of a trend taking off; the method is in [how to put yourself into any viral video](/blog/put-yourself-in-any-viral-video-with-ai).

## Peaking now

### Bulin 47 freestyle

A fan with a speaker stops a Dominican dembow artist outside a club and they trade bars on the spot; one line, "soy Bulin 47", became the punchline the internet puts in everyone's mouth. Spanish-speaking feeds first, then everywhere. One performer facing the camera, fourteen seconds, audio that does the work: the ideal remix. [Template and guide](/blog/bulin-47-freestyle-ai-trend).

### Migos — Hotel Lobby

Quavo and Takeoff in the orange COLORS booth, one hanging microphone, four medium shots. Two slots, so two of your characters take the parts and rap the verse with the same moves and the same lip-sync. The cuts are what makes it hard for a face swap and easy for video-to-video. [Template and guide](/blog/migos-hotel-lobby-ai-trend).

### Migos — Carpool Karaoke

The windshield camera, four seats, quick cuts between them on the "Walk It Talk It" hook. Up to four characters; the version with a football team reposted by a driver is why it came back. [Template and guide](/blog/migos-carpool-karaoke-ai-trend).

## Still running

- **Yung Lean storm dance** — the dance in the rain, one performer, eight seconds. [Guide](/blog/yung-lean-storm-ai-trend).
- **Ceiling fan outfits** — the rescue from the fan, with an outfit change per moment; the template takes a look per shot. [Guide](/blog/ai-ceiling-fan-trend).
- **Shadowboxing in Paris** — the shadowbox walk, one performer. [Guide](/blog/ai-shadowboxing-trend).
- **Smile, cry, walk away** — three expressions, one take. [Guide](/blog/ai-crying-trend).
- **Save yourself** — the cliff-edge rescue, two slots. [Guide](/blog/save-yourself-ai-trend).

## On the desk

Trends spotted this week that do not have a template yet. If one of them takes off, it gets one.

- **"Alors on danse" studio duet** — a short white-studio performance clip on the Stromae track, two performers at a keyboard and a laptop, recast versions passing hundreds of thousands of views on X in a week. Two slots, a fixed shot: a natural fit.
- **Rumpelstiltskin, "Tip Toein' In My Jordans"** — an AI-generated period-film clip of Rumpelstiltskin dancing for a princess on the Riff Raff track, with replacement versions everywhere. One performer, one dance.
- **"Ella y yo" recreation** — the Aventura and Don Omar duet staged with two football rivals, viral in Argentina after a famous repost. Two slots, a conversational staging, Spanish audio.

## How a trend becomes a template

A clip qualifies when it has one short window everyone reuses (five to fifteen seconds), a performer facing the camera, and audio that carries the joke. The template is that window, trimmed, with the slots named and the instruction written, so that making your version is: pick the template, cast your character, press Replace. Replace Pro renders it in about five minutes with the original sound, at 59 credits per second of clip, so a ten-second trend costs 590 credits. The identity sheet your character renders from costs 20 credits once.

The rule that applies to every template: your own face or your own characters. Never a real person who did not agree, never a minor. The [acceptable-use policy](/legal/aup) is enforced before a render is billed.

## FAQ

### How do I know which trend to make?

The one your audience already knows the audio of. A trend works because the viewer recognises the clip before they notice the face; a clip nobody has seen is just a video.

### How fast do new trends get a template?

Within a day when a trend qualifies: the desk checks every morning. Trends that need a text prompt only, image trends, and anything involving minors do not get one.

### Can I request a trend?

Yes: the support form in the studio, with a link to the original clip. The ones with a clear performer and a reusable window go first.

### Does the result have the original audio?

Yes, on Replace Pro: the source's sound is put back under the result, cut at the clip's length. Replace (the words-only engine) returns picture only unless you ask for the sound back.
