Hotel Lobby AI Trend: How to Make the Video (2026 Guide)

If your feed has been full of two people standing in a bright orange room, trading bars under a single hanging microphone — cats, athletes, cartoon characters, your coworkers — you've met the Hotel Lobby AI trend.

It's the biggest AI meme format of the moment, and the good news is that making your own takes about two photos and five minutes. The less obvious part is everything around it: why the shot looks the way it does, which tool to pick, what breaks your generation, and — the question almost every guide skips — what audio you're actually supposed to use.

This guide covers all of it.

What the trend actually is

First, the name is misleading. There is no hotel and there is no lobby. The name comes from the song.

Here's the actual origin, precisely:

In 2022, Quavo and Takeoff — performing under the duo name Unc & Phew — released "Hotel Lobby" on May 20. It was produced by Murda Beatz, Keanu Beats, and Fabio Aguilar, and released through Quality Control / Motown as the first single from their album Only Built for Infinity Links.

On June 17, 2022, the pair performed it on A COLORS SHOW, the Berlin-based music platform whose whole format is a solid-colour backdrop, one microphone, and no distractions. That performance — two rappers, one hanging mic, a flat orange wall — passed 112 million views, making it one of the three most-watched COLORS performances ever.

Then it sat there for four years.

In early September 2026, someone put two cats into that orange room. The videos went viral. Within weeks the internet was recasting the performance with everyone: Barack and Michelle Obama, Walter White and Jesse Pinkman, Spider-Man, Erling Haaland, people's parents, their dogs, their siblings. Thousands of them.

On September 23, 2026, Quavo reposted the untouched original performance with the caption "The original!!! #HOTELLOBBY." That reclaim pushed the trend into a second, larger wave — and it's still going.

Why the format works so well, if you're curious: the COLORS staging is deliberately minimal, so it reads instantly even with the sound off, and a stranger can stand in for either performer without breaking the joke. Simple stage, copyable gesture, two-person staging. That's a template engineered by accident.

How to make a Hotel Lobby AI video

Four steps. You need two photos — no prompt writing, no editing software.

Step 1: Pick your two subjects and get one photo each

You need two separate images, one person (or pet, or character) per photo. Not a group shot.

Your first photo becomes the performer on the left; the second becomes the right. Decide who goes where before you start, because that ordering is the one thing you can't fix after the fact.

What makes a good source photo:

• One subject only — crop a group photo down before uploading
• Face clearly visible: eyes, nose, mouth all unobstructed
• Front-facing or a slight three-quarter angle
• Even lighting — no harsh overhead shadows, no backlighting
• No sunglasses, masks, or heavy shadow across the face
• Full body if you have it (more of the body in frame means better full-body reconstruction)
• Sharp enough that you'd recognise the person without zooming in

Pets work especially well — a dog or cat standing upright next to its owner is one of the most-shared variations. Drawn characters work too, if you have a clear illustration.

Step 2: Upload them in the right order

Open your chosen tool's Hotel Lobby template and upload the left performer first, then the right. Check the previews before generating — those two slots decide who ends up on which side, and getting them backwards is the single most common mistake.

If the reference preview doesn't load, reload the page before adding your photos.

Step 3: Choose your resolution and generate

Most tools offer 480p, 720p, and 1080p. If you're still testing which photos work, start at the lower resolution — higher resolution adds detail, but it can't fix an unclear face or a distorted pose, so there's no point paying for it while you're iterating.

Uploading photos does not start a job by itself. You have to hit Generate. And if a job is already processing, check that entry before submitting again — a second click can create a second paid attempt.

Step 4: Watch the whole clip before you download

This step gets skipped constantly, and it's where most disappointing results are caught.

A good opening frame can hide problems later in the clip. Watch from start to finish and look for:

• A face changing during a turn
• An outfit switching mid-performance
• Limbs merging when the performers move close together
• Either subject drifting to the wrong side

Then listen with sound on. (More on that below — it's the part nobody warns you about.)

Which tool should you use?

The market moved fast here. As of late September 2026 there are at least eight or nine tools shipping a Hotel Lobby template, and they split into three genuinely different approaches — not just different prices.

Price spread is wide. Fuzana lists 12 credits for a 5-second clip and 24 for 10 seconds, with packs starting at \$4.99. Starrd is reported around \$5 for a 15-second vertical clip. Trendai's 30-second 720p render has been listed in the thousands of credits. Read the credit estimate next to the Generate button — screenshot prices go stale.

My practical recommendation:

• Just want to try it once, today? Take the one-click template. Fastest path to a result, and you'll learn whether your photos are good enough.
• Posting cross-platform or building a channel? Use a prompt-generated approach, so nothing of the original performance goes into your render. You skip an entire category of problem.
• Want the exact original choreography? Motion transfer, but budget real time for sourcing the clip and expect a few regenerations.

What variations are actually getting posted

If you want your version to travel, here's what's working:

• Pet + owner — a dog standing upright beside its human is the single most reliable version. The trend literally started with cats.
• Best friends or siblings — the default, and still the best one
• Parent + kid / baby versions — high share rate, very low cringe
• Couples — anniversary and date-night posts
• Characters or fictional pairings — Spider-Man, Breaking Bad, cartoons
• Celebrities and athletes — the most popular and the most legally exposed; see the rights section below
• The unexpected pairing generally — the joke lands hardest when the two subjects clearly don't belong in a rap booth

One structural tip: whichever variation you pick, the format depends on a back-and-forth. You want one performer leading while the other reacts, then swapping — not both doing the same gesture on the same beat. Independent movement is what makes it feel conversational, and it's also easier for the model to render accurately.

Troubleshooting the common failures

Faces merging into each other Usually a photo problem, not a tool problem. Use two clearly separate photos rather than one group shot, and make sure there's a visible gap between the two bodies in the frame. Promote prompts that ask for a clean gap and no crossing hands.

Subjects switching sides mid-clip Fix the left/right assignment in your prompt explicitly — state that the first photo is on the left and the second on the right, and that they never swap. Some tools also let you re-assign roles; use Replace on the affected slot rather than regenerating everything.

The clip drifts or pushes in Ask for a locked-off camera, one continuous take, no cuts, zooms, or pans. Camera drift is the fastest way to make a booth clip feel like a different video.

It doesn't look like the person The photo is the problem 90% of the time. Sunglasses, heavy shadow, backlighting, and group shots are the top four causes. Front-facing, evenly lit, one person, face fully visible.

Regenerate the smallest failing detail rather than adding more instructions. Piling on prompt text usually makes the whole thing worse.

The part nobody covers: the audio

Here's where every other guide stops and the real problem begins.

Most templates keep the original performance audio underneath your face swap. That's what makes the result feel instantly recognizable — and it's also the part that creates a problem the moment you post it somewhere other than where it's licensed.

The short version:

• Inside TikTok or Reels, attached as the official sound from the app's library, in-app use is covered by those platforms' licenses. This is fine and legitimate.
• The moment the audio leaves the short-form app — a YouTube upload, a cross-post, a long-form video, a livestream, an ad, client work — that license doesn't follow. YouTube's Content ID runs audio fingerprinting, and since 2025 it also runs AI music detection that catches pitched-down, sped-up, and filtered versions of registered songs.
• The original is a real recording with real rights holders: the label, the publishers, and the producers. Not a gray area you can outrun with a pitch shift.

And there's a second consideration that isn't about law: Takeoff passed away on November 1, 2022, four months after the COLORS performance was filmed. His likeness being recast in these memes has drawn real criticism, and it's worth being thoughtful about rather than treating it as free material.

So what do you actually use? There are three legitimate answers — post natively in-app with the official sound, score it with an original or royalty-free beat, or build your own audio identity into it. I wrote the whole thing up separately, including what the original beat sounds like if you want to match its energy:

→ Hotel Lobby AI Trend Audio: 3 Legal Ways to Fix the Sound

The short version of the third option: if you're going to keep making videos, make your own tag and put your name on them. You can generate a producer tag free in about ten seconds — no login, no recording gear — or clone your own voice if you want it unmistakably yours. That audio works everywhere, permanently, because it's yours.

Rights, likeness, and staying out of trouble

Being straight with you, because this is the part that quietly ruins people's channels:

Don't post celebrity face swaps commercially. Making one for a laugh is one thing. Putting Obama, a footballer, or a studio character into anything you're paid for, whitelisting, or running as an ad is a different category of risk entirely. The AI platforms are already steering users away from this.

Only upload photos of people who agreed to be in it. Handing a stranger your photos also hands them your face. Privacy policies on these tools vary.

Don't use the original recording outside the licensed in-app context. Covered above, but it bears repeating, because it's the single most common way this trend turns into a takedown.

Disclose AI use where it's appropriate — several platforms now auto-label photorealistic AI content anyway, and YouTube began auto-applying AI labels in 2026.

Build your own duo. The trend works just as well with you and your best friend as it does with two celebrities, and it carries zero rights exposure. The joke is the format, not the famous faces.

Frequently asked questions

Is the trend set in a hotel lobby? No. The name comes from the song "Hotel Lobby." The visual is Quavo and Takeoff's A COLORS SHOW performance: a flat orange backdrop, one hanging microphone, two performers, no set dressing.

Who are the original performers? Quavo and Takeoff, performing as Unc & Phew. The COLORS performance was filmed June 17, 2022. Takeoff passed away on November 1, 2022.

How many photos do I need? Two — one person per photo. A group shot gives the model one crowded read and the identities blur.

Can I use pets or cartoon characters? Yes, and pet versions are among the most-shared. A drawn character needs a clear, unobstructed illustration.

What song is it? "Hotel Lobby" by Unc & Phew, released May 20, 2022, produced by Murda Beatz, Keanu Beats, and Fabio Aguilar.

Can I put myself on both sides? Some tools allow it, but the format relies on two distinct identities trading verses. Same-face versions tend to look uncanny.

How much does it cost? Templates run a few dollars per clip — Fuzana starts at \$4.99 for credit packs, Starrd is reported around \$5 per 15-second clip. Resolution and clip length drive the cost. Read the credit estimate before generating.

Why does my result look nothing like the photos? Almost always the source images. One subject per photo, front-facing, even light, no sunglasses or masks, face fully visible. Retake the photo before you retry the tool.

Is the trend still going? Yes. It started in early September 2026, got a second wave when Quavo reclaimed the original on September 23, and remained active through the end of the month. Formats like this typically run for weeks to a few months, though how-to content keeps drawing long-tail traffic long after the peak passes.

The short version

Two photos, four steps, a few dollars. Pick a tool based on whether you want speed, exactness, or zero reuse of the original. Watch the whole clip before you post it.

And decide your audio before you publish, not after — that's the step everybody skips, and it's the one that determines whether your video can live anywhere other than the app you posted it in.

The original A COLORS SHOW performance of "Hotel Lobby" by Unc & Phew is available on COLORS' official channel. Quavo and Takeoff released the track on May 20, 2022; the COLORS performance followed on June 17, 2022. Takeoff died on November 1, 2022, at 28.