Hotel Lobby AI is a trend in which two characters, often pets, rap back and forth under one hanging microphone in a bare orange studio, and you can make your own from two photos on the Hotel Lobby AI trend page. The result is a silent 15-second video, so you add the original sound in TikTok or CapCut.
This guide goes past the basics and answers the questions that come up once you start making one: why this clip, how to sync it, how to avoid face mix-ups, and how to handle the Takeoff conversation with care. The sources are Complex and Revolt, plus the videosi page itself.
Key takeaways
- The clip is silent and not an exact lip-sync; add the Hotel Lobby sound yourself and align the first bar with the cut at about 2 seconds.
- Two clear, separate, front-facing photos beat any clever description, and the prompt is built in.
- Turn on the AI-generated label and keep the original performers out of your remake.
- A 15-second Standard 720p video costs 13 credits, or 7 at 480p, and failed renders are refunded.
In this guide
- Why did the COLORS performance become the template?
- Why are the pet versions the most shared?
- How do I line the clip up with the song beat?
- Is it a true lip-sync of the song?
- How do I stop the two faces from blending?
- Should I label the video as AI-generated?
- What about copyright and using the original song?
- How is this different from a face-swap version?
- How should the Takeoff discussion shape what I make?
Why did the COLORS performance become the template?
The COLORS clip works as a template because almost nothing in it is hard to copy. Complex describes the original as a 2022 A COLORS SHOW performance of "Hotel Lobby" by Quavo and Takeoff, billed as Unc & Phew, and Revolt notes it was filmed for Black Music Month. The set is a plain colored backdrop with one microphone, two people, and a back-and-forth delivery.
That simplicity is what an AI model needs. There is no crowd, no band and no moving set to rebuild. A generator only has to keep two characters recognizable, hold one flat orange look, and repeat a clear rhythm of pointing, leaning in and turning away. The Hotel Lobby AI page bakes this scene in, so you never describe the studio yourself.
The format is also easy to read at a glance. Viewers know within a second what they are looking at, and the joke lands before the first bar. That is why so many unrelated pairs, from cats to couples, could fit the same frame.
Why are the pet versions the most shared?
Pets are the funniest mismatch with the swagger of the original. Complex reports the trend began in early September 2026 with videos of two cats recreating the performance, and only afterwards did people swap in celebrities, fictional characters and athletes.
A cat standing upright and gesturing like a rapper is absurd in a way a face swap of a famous person is not, and it carries no consent problem. Revolt also mentions the Obamas, Beyoncé and Blue Ivy and Woody and Buzz Lightyear among the later swaps, so the format spread quickly beyond animals. Pets still have one advantage: nobody objects to them.
On videosi the pet rule is built in. Each animal keeps its species, coat and markings and stands on its hind legs, with its front paws doing the gesturing. For best results use a full-body photo where the markings are visible, one animal per photo.
How do I line the clip up with the song beat?
Add the sound in TikTok or CapCut and match the first bar to the cut. The videosi clip is a silent 15-second MP4, so the audio is always your job, and the one landmark you can trust is the cut from the wide shot to the two-shot at about 2 seconds.
A practical routine looks like this:
- Import the clip and the Hotel Lobby sound into your editor.
- Place the start of the first bar on the cut at roughly 2 seconds, then nudge by a frame or two until the lean toward the lens feels on the beat.
- Trim the sound so that the last few seconds land on the finger-to-lips ending.
Because the characters gesture in rhythm rather than sing the words, small drift is forgiving. If a section feels late, move the audio rather than the video. The full beat sheet is on the trend page.
Is it a true lip-sync of the song?
No. The result is a performance in rhythm, not an exact lip-sync. The mouths move as if rapping, the hands and heads follow a beat, but the model renders picture only and the file has no sound at all.
This matters for expectations. Tools that map two photos onto the original footage can keep the original audio and mouth shapes, as Complex describes for some apps. videosi takes a different route: it generates new footage of your two characters, borrowing only timing and body rhythm from a reference performance, never anyone's face. The trade-off is that individual syllables will not match.
In practice viewers rarely check syllables. They watch the swagger, the pointing and the turn-away. Say clearly in your caption that the clip is a remake, and the lack of exact sync reads as part of the joke.
How do I stop the two faces from blending?
Give the model clean, separate evidence for each performer. The most common causes of blended or swapped faces are tight face crops, sunglasses, dark or blurry photos, and group photos with more than two faces.
The simplest fix is two photos, one per performer, each front-facing and lit evenly, showing shoulders and outfit. Photo 1 becomes the left rapper and photo 2 the right one. If you only have a shot of both of you, use the one-photo mode and stand side by side with nobody else in frame.
Make the two characters look different from each other. Two people in near identical dark jackets and caps are the hardest case. A different top color or a different hair shape helps the model keep them apart through the turn-away and spin back. Details are on the Hotel Lobby AI page.
Should I label the video as AI-generated?
Yes, turn on the AI-generated label when you post. It is a one-tap setting in TikTok, and it keeps you on the right side of platform rules and your audience's trust.
The trend itself drew criticism. Complex reports fans objecting to the use of Takeoff's likeness after his death and wider worries about consent and how AI platforms handle uploaded photos. Labeling your post will not answer those worries, but it avoids the extra problem of passing synthetic footage off as real.
It also helps with search and recommendation. Viewers who browse the AI tag are the ones looking for exactly this kind of remake, and a clear caption such as "AI remake with my cats" sets the tone before they press play.
What about copyright and using the original song?
The song is Quavo and Takeoff's work, so the audio belongs to its owners and to the platform's licensing terms. videosi does not supply or embed it, which is why the clip is silent.
Using the sound through TikTok or CapCut's own library is the normal route, because those apps handle licensing for in-app use. If you post elsewhere, check what that platform allows before you add music. Commercial use is a separate question, and a brand ad built on the song needs real permission, not just a trending sound.
The same logic applies to likeness. Revolt reports that Quavo posted his own non-AI recreation with Victoria Monét on 1 October, captioned "No AI", which shows the original artist is happy to join the format on his own terms. Your remake should stay a clearly labeled fan creation, and it should not imply that anyone endorsed it. videosi is independent and not affiliated with TikTok, CapCut, COLORS or the artists.
How is this different from a face-swap version?
A face swap pastes your faces onto the original footage, while videosi generates new footage of your characters. That one difference changes the audio, the likeness issue and the privacy picture.
With a swap, the body, the room and the original performers' movements stay in the frame, so the audio and lip movements can match. It is also the version that raised the loudest objections about Takeoff, since his actual performance sits underneath. Complex describes tools that map two uploaded images onto the original video.
With the videosi approach, only timing and body rhythm come from the reference. Your two characters, their clothes and the orange studio are produced fresh, and the clip is silent. You lose exact sync, but nobody's real footage is reused. If you want to compare it with the full explanation of the trend, start at the Hotel Lobby AI trend page.
How should the Takeoff discussion shape what I make?
Keep it respectful and keep the original performers out of it. Takeoff was killed in Houston on 1 November 2022, according to Complex, and some fans see AI versions that reuse his likeness as upsetting.
videosi handles this at the door. It does not accept photos of celebrities, public figures or minors, and that includes Quavo and Takeoff. Use photos of yourself, of adults who agreed, or of your own pets. Users must be 18 or older, and photos and results are deleted within 30 days.
You can follow the same spirit in your caption. Skip jokes about the original duo, credit the song, and let the cats or friends be the joke. Quavo's own response, per Complex, was to repost the untouched COLORS video on 23 September with the line "The Original". Pointing your viewers toward the original is a fair way to honor it.
Sources
Background: Complex on the Hotel Lobby AI trend and Revolt on the Victoria Monét and Quavo version.
