Hotel Lobby AI Video: The Duo Rap Clip
Upload a photo of each person, and the AI puts both faces on the two performers in the booth — one trading bars with the other, mouths synced to the vocal, in the landscape booth framing of the source clip.
Credits to start · No filming · Crop for Reels, Shorts and TikTok after export
Reference performance
The Hotel Lobby preset is loaded by default. Remove it to use your own 3–15s motion reference.
First person
Required · left performer, or a photo containing both subjects.
Second person
Optional · right performer when using separate photos.
The Clip You Get
A short video of two people performing in the lobby booth: both faces replaced, both mouths moving on the beat they are assigned, cut in the landscape 16:9 framing of the source performance.
The scene itself is fixed — the room, the light, the microphone stay exactly as they are in the source clip. What changes is who is standing in it. That is the whole appeal of this format: viewers recognise the room instantly, then spot that it is their friends inside it.
Most versions of this video floating around are landscape, because the clip was shot for a wide screen and that is how it gets passed around first. Cropped to 9:16 without thought, one of the two performers usually falls off the edge. The framing section below is about not letting that happen when you post tall.
Two Photos, Two Performers, One Booth
1. Pick two photos that can sit in the same room
Both faces should be shot under similar light and from a similar angle. A daylight selfie next to a photo taken in a dark bar is the single most common reason a duo swap looks fake: the model corrects exposure per face, and the two corrections do not match.
2. Assign each face to a performer
The uploader asks which photo goes on the left performer and which on the right. This is not cosmetic — it decides who mouths the first bar. Get it backwards and the whole opening verse lands on the wrong person.
3. Generate, then frame it for where you post
Output follows the booth template in landscape 16:9. If you need tall for Reels, Shorts or TikTok, crop or letterbox after export — regenerating will not change the scene’s native framing.
Why This Scene Reads as Rap
The booth is built like a rap performance, not a photo set. Two people stand at a shared microphone rather than posing at separate marks, the framing is a stage wide shot, and the lighting is coloured and directional instead of flat. That combination is what makes a replaced face read as "performing" rather than "photoshopped in".
It also explains why the format took off with rap audio specifically. The performers in the source clip are already delivering bars with their heads moving and their mouths working — which is exactly the motion a face swap needs in order to look alive. A scene of two people sitting still would map a static face onto a static body, and the illusion collapses.
Getting the Booth into 9:16 Without Losing a Performer
This is the part nobody covers, and it is where most vertical versions of this video go wrong. The booth is a wide composition: one performer sits toward the left third of the frame, the other toward the right, and the microphone hangs between them. A straight 9:16 crop out of the centre keeps the mic and cuts both performers in half. This page exports the wide shot; the work of going tall happens when you post.
Three ways to handle it, in order of how good they look:
- Crop with both performers in mind. Pull a tall window that still includes both heads and the mic, even if you lose side wall. Better than a dead-centre crop that halves both faces.
- Pan between the two performers. Keep 16:9 and cut a moving tall window that follows whoever is rapping, switching on the beat. Effective, and it hides the crop, but it needs a cut per bar and it looks busy if the switch is off by a beat.
- Letterbox the wide shot. Blur a copy of the frame behind a centred wide video and stack it in a tall canvas. Safe and quick, and it clearly reads as a recycled landscape clip — which is exactly what it is.
If you only post to one place, pick the ratio of that platform when you edit. Reels, Shorts and TikTok all want tall; a wide version has to be reworked before it will perform there.
Which Face Gets Which Bar
In the source performance the two parts are not equal. One performer opens and carries the first bar, the other answers. If you want the output to sound like it was made for the two people in the photos rather than stamped onto a template, match the assignment to their relationship.
Two people who actually know each other: give the opening bar to whoever talks more. The answer bar reads as a reaction, and a reaction only lands if the viewer believes the other person would react that way.
A couple: the answering part usually gets the better line, so put it on whoever is more likely to have the last word. This is why so many of these videos are funny — the swap reverses who wins the exchange.
If you want to change who speaks first, swap the two uploads rather than editing after generation. The lip sync is baked into the render, so the assignment has to be right before you press generate.
The Audio Question: the Original Track or Your Own
Every version of this trend you have seen runs on the same audio, and that audio is a copyrighted commercial recording. That matters more than it looks:
- Posting it on a social platform generally falls under that platform's music licensing for organic posts, and the platform may mute, limit or share revenue on the post. That is not a universal commercial licence.
- Using it in an ad, a brand video, or anything you are paid for is a different situation entirely and needs a licence you probably do not have.
- Re-uploading the clip with the original audio to a site you monetise is the riskiest use of the three.
The clean route is to generate or upload your own track and swap that in. The booth scene does not care what is playing — the mouth movement is generated against whatever audio you give it — so a track you own produces the same format with none of the exposure. If you do want the original feel, ask for the same tempo and delivery rather than the same recording.
Captions and the Loop
Two details separate a clip that gets shared from one that gets scrolled:
Word-by-word captions. A single line of text per bar, appearing as it is delivered, sitting in the empty booth space above the performers' heads. This is the one part of the frame that is always safe to cover, and it is why a careful tall crop can be an advantage rather than a compromise.
A chorus that loops. If the clip ends on the hook rather than on a cold last bar, the last frame flows back into the first and viewers watch it twice without noticing. Keep the export length at the length of the loop.
Two Faces That Look Like They Belong Together
The difference between a swap that works and one that gets commented on for the wrong reason is almost always the pair, not the model.
- Same light. Both photos in daylight, or both indoors under the same kind of lamp.
- Same angle. Both facing the lens. One profile shot will break the pairing even if everything else is perfect.
- Same crop. Hairline to chin for both. If one photo is cropped at the forehead, the two faces will sit at different heights in the booth.
- Similar sharpness. One soft, filtered photo next to a crisp one reads as two different videos stitched together.
- No sunglasses, no hats pulled low, no heavy beauty filters.
If the result looks off, reshoot the weaker photo before you regenerate. A better photo fixes more than any setting will.
Other Scenes Worth Trying
The same two-photo flow works on every scene in the library:
Night stage
Darker, heavier contrast, better for drill and trap.
Rap battle
The two performers face off instead of sharing a mic.
Cypher
A group scene where you take the front two positions.
Solo booth
One photo, one performer, if you only have one face to work with.
Bulin 47
Another duo energy with its own staging.
Burning Bridges
A different performance cut for a second take.
Rumpelstiltskin
Hay-barn tip-toe clip with film grain intact.
FAQ
How do I make a Hotel Lobby AI video with two people?
Upload one photo of each person, assign them to the left and right performers, and generate. The AI replaces both faces in the booth scene and syncs each mouth to the bar that performer delivers.
Can I make it vertical for TikTok, Reels and Shorts?
This scene exports landscape 16:9 to match the booth template. For tall feeds, crop or letterbox after download — a centre crop usually cuts a performer off, so pan, letterbox, or crop carefully so both stay in frame.
Do I need two photos, or does one work?
Two is the standard version, and it is what the scene is built for. If you only have one usable photo, use a solo scene on the homepage instead of leaving the second performer unchanged.
What if only one of my two photos is good?
Replace the weaker one first. One underexposed or angled photo is the most common reason a duo swap looks pasted on, and no setting compensates for it.
Can I change who raps first?
Swap the two uploads and regenerate. The lip sync is part of the render, so the assignment has to be correct before generating rather than fixed afterwards.
Can I use my own audio instead of the original track?
Yes, and for anything commercial you should. The scene works with any track — the mouth movement is generated against whatever audio you supply. Platform music libraries may licence a track for organic posts; that is not a blanket commercial licence.
Is the Hotel Lobby AI video free to make?
Generation uses credits. You can load the preset and choose photos before spending credits; pricing is on the pricing page. Exports do not carry a watermark.
Can I put pets or characters in the booth instead of people?
The swap needs a face with visible eyes and a mouth to track. A clear pet portrait can work; a mascot with a fixed plastic expression usually does not.
Make Your Hotel Lobby Rap Video
Upload two photos. Assign left and right. Generate the duo clip when you are ready.
Generate the duo clip