How to make a Hotel Lobby AI video with two photos
Updated 15 October 2026
A "Hotel Lobby" AI video is a short clip in the same orange room as the 2022 live performance this trend is built on, with the two performers replaced by the two people in your photos. There is no prompt to write and no editing to do — you upload one photo per person and press Generate. This page is the whole job, in order, including the parts that can go wrong.
What the clip actually is
The room, the camera and the performance all come from one existing live session — the orange studio with the microphone hanging from the ceiling. The trend works because the room never changes: the setting is the joke, and the two faces inside it are the punchline.
So the tool has one job. It takes the room and the performance from that footage, and it takes the two identities from your photos. It is not a filter and not a template you drag photos into; the clip is rendered, which is why it takes minutes rather than seconds.
Before you upload anything
Two things matter more than anything else you will do on this page: the quality of the photos, and whether you are allowed to use them.
- You need two photos — one per person, one face per photo.
- Only upload photos of people who have agreed to appear. Uploading someone without their consent breaks the acceptable-use policy, and it is the quickest way to have a clip taken down.
- Do not use photos of public figures or anyone you are trying to impersonate. Clips that impersonate real, identifiable people are refused.
- No nudity, no sexualised content, nothing involving minors.
The seven steps
- 1Open the generator. It is on the home page — the two empty slots next to the finished clip.
- 2Put one person in LEFT and the other in RIGHT. The left slot sings the opening, so put whoever should start on the left.
- 3Upload. Each photo is checked for type and size before anything is stored; a rejected file costs nothing.
- 4Wrong sides? Press Swap. It exchanges the two people without re-uploading anything.
- 5Choose the clip. Length 5, 10 or 13 seconds; quality 768P or 2K; shape 4:3, 16:9, 9:16 or 1:1. Every shape costs the same.
- 6Press Generate. This is the only step that spends credits — browsing and uploading are free.
- 7Wait, then download. Rendering takes a few minutes and continues if you leave the page. Download links stay live for 7 days.
Which photo to pick
The model copies identity and wardrobe from your photo and the performance from the room. A photo that does not clearly show a face gives it nothing to copy, which is the most common reason a result disappoints.
- One person per photo. Group shots and photos where a second face is visible cause the two identities to blend.
- Face the camera. A three-quarter or profile shot still works, but a straight-on photo works best.
- Even light, no sunglasses, no mask, no heavy shadow across the face.
- Head and shoulders is plenty — this is not a full-body shot, and the clip is framed from the waist up.
- A sharp photo beats a flattering one. Soft or heavily filtered images lose the details the render needs.
Choosing length, quality and shape
Length is the main decision, and shorter is not simply cheaper — it changes which part of the performance you get. The performance is about 13 seconds long, so 13 seconds is the only setting that contains the whole thing, including the turn near the end. Five and ten seconds buy the opening of the same performance.
Quality: 768P is the standard tier and what most people should use. 2K costs more credits and is worth it if you plan to watch the clip on a large screen.
Shape: 4:3 matches the original footage most closely. 9:16 is the one to pick for TikTok, Reels or Shorts. 16:9 suits YouTube or a desktop player. 1:1 suits a feed post. All four are the same price, so pick the shape you will actually publish in.
If the faces cannot be verified
Some photos are refused — usually because the face is too small, too blurred, or at an angle. When a clip fails for that reason the credits are returned automatically; you are not charged for a clip you never receive.
If your first attempt fails, upload a different photo rather than pressing Generate again on the same one. Re-running identical inputs tends to produce an identical refusal.
What it will not do
Knowing the limits saves you credits. This tool follows the performance; it does not reproduce it frame for frame, and it will not turn two photos into an arbitrary scene.
- It follows the reference performance rather than copying it exactly. Expect the gestures and the timing to match, not every pixel.
- Mouth movement is generated to fit the performance. It looks natural in motion; it is not a precision lip-sync to any particular recording.
- Exactly two people. There is no way to add a third performer.
- 13 seconds is the ceiling, because that is how long the performance is.
- You cannot bring your own video or your own script. The room and the performance are fixed on purpose — that is what makes the clip readable as the trend.
- No public figures, no impersonation, no NSFW.
Questions
Do I need to write a prompt?
No. The instruction is written for you. See the prompt page if you want to know what is actually being sent on your behalf.
How long does it take?
A few minutes. Rendering runs in the background, so you can close the tab and come back — the clip is waiting on your clips page.
How much does one clip cost?
It depends on length and quality: 12 credits for a 5-second 768P clip, up to 45 credits for a 13-second 2K clip. Shapes are free. A starter pack of 50 credits is $8.90, and credits do not expire.
Do I need an account before I can try it?
No. You can open the generator, upload two photos and see the whole flow without signing up. The account is only required at the moment you press Generate.
Can I use the clip commercially?
You are responsible for having the rights to the photos you upload. Whether you may use the result commercially depends on your own situation and the music, so check the terms page before publishing anything you intend to monetise.
Make one with your own two photos
Uploading and previewing are free — you only spend credits when you press Generate, and a clip whose faces cannot be verified is refunded automatically.
Face can’t be verified? Credits are refunded.