AI Hotel Lobby Video: Recreate the Two-Person Booth Trend From Two Photos
Two people, one hanging microphone, an orange booth. The format is simple enough to be recognised instantly and specific enough to be hard to generate well — the failure is almost always the same one: the two faces swap sides halfway through.
The Hotel Lobby trend comes from a 2022 COLORS performance by Unc and Phew — Quavo and Takeoff — shot as two performers trading verses under a single hanging microphone in a bare orange booth. In September 2026 it became the most remixed clip on short-form feeds, with thousands of versions swapping the two original performers for anyone with a photo: couples, parents and children, pets, fictional characters.
What makes it difficult is not the setting. It is that there are two identities on screen at once and the model has to keep them on the correct sides through every cut. Get that wrong and the result is a clip where faces blend or swap — which is the single most reported failure in every tutorial on the trend. OpenFilmo approaches it by making the left and right assignment explicit at upload time rather than leaving it to the prompt.
What it does
Built for this specific job, not a generic generator.
Two photos, one clip
Upload one front-facing photo per performer. No source video to find, no template licence to buy, no clip of the original performance to trim.
Lock each performer to a side
The first photo is left, the second is right, and that assignment is held through every cut instead of drifting mid-clip.
Vertical first
Export 9:16 for TikTok, Reels and Shorts, or 1:1 and 16:9 from the same generation.
Finish in the same tab
Captions, music, voiceover and a 4K upscale are in the same workspace, so the clip is finished rather than half-made.
How it works
Four steps, no timeline to learn.
Choose two clean, front-facing photos
One person each, good light, no sunglasses, face clearly visible. A pet or a drawn character can take one side. Avoid heavy backlight and profile angles — the model needs to see the face.
Decide who stands left before you upload
This is the step people skip. The first photo you upload is the left performer and the second is the right. Decide it first, then upload in that order.
Generate a short test clip
Render five to ten seconds at lower cost first and check both faces landed on the right side. Ten seconds tells you whether the assignment held without spending a full-length render on it.
Render full length, then finish
Once the test holds, generate the full clip. Add captions and audio in the same workspace, then export per channel.
What people make with it
The jobs this actually gets used for.
Couples and weddings
Two people who would never normally be in the same video booth, trading verses to camera.
Parents and children
A parent and a child on either side of the mic — among the most rewatched versions of the format.
Person and pet
A dog or cat taking one side. Animals hold up well because the model only needs to keep one face stable.
Teams and brand accounts
Two colleagues or two products personified, used as a launch or announcement clip.
Why two-person trends need explicit assignment
The setting is easy. Keeping two identities apart is not.
Single-subject trends are forgiving: one face, one outfit, one side of the frame. The moment a second identity appears, the model has to solve an assignment problem across every cut, and most failures reported by creators are exactly that — faces smearing into each other, or the left performer appearing on the right halfway through.
The practical fix is to remove the ambiguity before generation rather than to prompt harder. Uploading in a fixed order and testing short means a failed assignment costs a five-second render instead of a full clip, and it also tells you which single variable to change on the retry.
There is also a rights boundary worth stating plainly. The format originates in a copyrighted performance by identified artists, and some platforms refuse uploads that reuse the original audio or staging outright. Generating a new booth from your own photos is a different act from reusing their take — check the platform rule before you post, especially for anything commercial.
Credits and plans
Generation is metered in credits, shared across every tool.
A five-second video costs 5 credits and one image costs 2. The Free plan gives you 30 credits every month with no credit card, no watermark and commercial use included. Paid plans start at $9 for 400 credits and add faster generation, parallel tasks and higher export resolution.
Related guides
Longer reads on the same workflow, from the OpenFilmo blog.
AI Video Prompt Formulas That Hold Up in Production
Prompt structures that work for AI video: subject, motion, camera, time and style — with tested formulas, failure patterns and a m…
Character Consistency in AI Video: What Actually Works
How to keep characters consistent across AI video shots: reference images, prompt discipline, key frame locking and post-hoc corre…
Image to Video AI: Animate Any Photo in Seconds
A complete guide to image to video AI: what makes a good source image, how to write motion prompts, top use cases, and how to fix …
AI tools
Single-purpose tools that run in the same workspace, on the same credit balance.
Vidu models
The Vidu Q1 and Q2 models this site runs on, and how they compare to Kling, Veo and Runway.
More OpenFilmo tools
Other generators and utilities in the same workspace.
Related workflows
Closely related tools and guides on the same problem.
Frequently asked questions
What is the Hotel Lobby AI trend?
A short-form remix of Unc and Phew performing Hotel Lobby for COLORS in 2022: two people in a bare orange booth under one hanging microphone, trading verses to camera. From September 2026 it became widely remixed with AI-generated faces.
How many photos do I need?
Two, one per performer, front-facing and well lit. The upload order matters: the first photo becomes the left performer and the second becomes the right.
Why do the two faces swap or blend?
Because the side assignment was never made explicit. The model has to decide which identity goes where, and it can change that decision between cuts. Assign left and right before generating and test with a short clip first.
Can I use a pet or a drawn character?
Yes. One side can be an animal or an illustrated character. It usually holds up better than a second human face, because only one identity has to stay stable.
How long should the clip be?
The widely shared versions run roughly fifteen to twenty-five seconds. Test at five to ten seconds first to confirm both faces landed correctly, then render full length.
Is the original song included?
No. Generate the visuals and add audio yourself in the workspace, using a track you have the rights to use on your platform. Reusing the original recording is a separate rights question.
How much does a Hotel Lobby clip cost?
A five-second video costs 5 credits. The Free plan gives 30 credits every month with no credit card, so a short test clip costs nothing. Paid plans start at $9 for 400 credits.
Can I use the result commercially?
OpenFilmo exports carry commercial use rights on every plan including Free. That covers your generated footage — it does not grant rights to any third-party audio or to the original performance this format derives from.
What aspect ratio should I export?
9:16 for TikTok, Instagram Reels and YouTube Shorts. 1:1 works for feed placements and 16:9 for anything embedded in a page.
What if the first attempt gets the sides wrong?
Change one thing only: swap the upload order. Changing the photos, the prompt and the length at once means you cannot tell which change fixed it.