Hotel Lobby, On The Radar and the New Wave of “Put Me in the Performance” AI Videos

For most of the past year, AI video trends looked the same: a prompt, a strange-looking clip, a few seconds of novelty. This autumn the most-shared clips on TikTok, Reels and X are different. They take a real, famous performance and swap the people in it. Two trends lead the way: the Hotel Lobby clip from Quavo and Takeoff’s 2022 COLORS session, and the “679” freestyle from On The Radar. In both, the original moves, the camera work and the audio stay exactly where they were. Only the faces change.

Why these two clips?

Both performances are short, rhythmic and filmed with almost no cuts. Two performers stand side by side at the microphones, trade lines and react to each other. That makes them easy to read in a vertical feed and easy to recast: friends, coworkers, pets and cartoon characters all look funny dropped into a booth that millions of people already recognise. The Hotel Lobby clip has its bright orange COLORS booth; On The Radar has its green-lit studio and neon sign. People know the setting in the first frame, so the joke lands immediately.

Search data shows how quickly this moved. Queries such as “hotel lobby ai” and “migos ai” climbed within days in late September, and “on the radar ai” — which first peaked in the spring when the 679 performance was posted — picked up again as creators looked for the next booth to recast.

How the swap actually works

The early versions of the trend were made by hand: generate a still of two people in an orange room, animate it, then lay the song on top. The results looked close, but the moves were invented, so mouths and gestures never matched the track. The newer tools take a different route. They start from the original performance and use video editing models to replace only the two performers, frame by frame, while keeping the choreography, the camera moves and every cut. Because the motion is the original motion, the original audio lines up on its own.

Tools built specifically for this format have appeared, such as a Hotel Lobby AI video maker where you upload two photos — one for the left performer and one for the right — and pick either the 15-second hook or the full 30-second chorus. The same approach now covers the green studio as well: an On The Radar AI video tool swaps two faces into the 679 freestyle. Both offer a cheaper quick-preview tier and a higher-quality tier that follows the original performance more closely.

What makes a good result

1. Clear, front-facing photos. One person per photo, both eyes visible, waist-up or full body so the outfit carries over.
2. Avoid sunglasses and heavy filters. The model has to rebuild each face through fast turns and close-ups; the clearer the source, the closer the match.
3. Left and right matter. The first photo replaces the performer on the left, the second the one on the right.
4. Ask before you post someone else. These clips spread fast. Use your own photo, friends who agreed, or characters you created.

Why it matters beyond the meme

The interesting part is not one song. It is that video models are now good enough to edit a real performance instead of guessing one. That opens a template economy: any short, recognisable clip with a fixed camera can become a format that people step into. Expect the next wave to come from other iconic booths, dance breaks and movie scenes. Platforms such as CreateVision AI are already turning these one-off trends into repeatable tools, and fans are treating them the way they once treated filters — something you try, share with friends, and move on from when the next performance takes over the feed.