Hotel Lobby AI Video: How to Make the Viral Migos Trend With Your Own Photos

What the Hotel Lobby AI video trend is

The Hotel Lobby AI video trend takes Quavo and Takeoff's "Hotel Lobby" performance on A COLORS SHOW — two performers in a bare burnt-orange booth with one microphone hanging between them — and swaps both of them for someone else. People use themselves and a friend, a couple, two pets, cartoon characters or celebrities.

• What makes it recognisable is the choreography, not just the song: the forward point, the forearm across the face, the turn away and the hush gesture, in the same camera framing as the original.
• Most versions use the 15 seconds that end on the line "in the hotel lobby". A longer cut runs the whole chorus.
• The original audio stays under the video, so the new performers look like they are rapping the real track.

That is why "make a dancing video from two photos" is not the same thing. A Hotel Lobby AI video has to keep the original performance moment by moment and only change who is performing it.

How to make a Hotel Lobby AI video in 3 steps

On CreateVision the source clip, the prompt and the audio are already set up, so you only bring the two people.

1. Open the Hotel Lobby AI video generator and add two photos — the first replaces the performer on the left, the second the one on the right. You can also pick a preset character for either side.
2. Choose the 15-second hook or the 30-second chorus, and a quality tier (the button shows the exact credit price before you start).
3. Click Swap Us In. The video takes a few minutes; you download an MP4 with the original audio mixed back in.

If you would rather run the model yourself, the prompt we use is further down. The rest of this guide explains the choices that decide whether the result looks like the trend or like a random dance video.

Which clip to use: the 15-second hook or the 30-second chorus

There are two cuts of the performance, and the choice changes both the result and the price.

• The 15-second hook is the version almost everyone posts. It ends on "in the hotel lobby", so the punchline lands at the end of the clip and it loops cleanly on TikTok and Reels.
• The 30-second chorus starts at "Water on me like the sauna" and includes every camera change. Use it when you want the full moment rather than the meme.

Quality tiers on CreateVision

Credits are held when you start and only charged when the video is delivered. If a swap fails, the hold is released automatically.

Photos that work (and photos that fail)

The photos decide more of the result than the model does. Each photo is the only information the AI has about that person's face, hair, outfit and body.

• ✅ One person per photo. Two photos, two people.
• ✅ Facing the camera, evenly lit, face and outfit clearly visible. Head-and-shoulders works; full body is better because the performance shows arms and torso.
• ✅ The outfit you want in the video. Whatever the photo shows is what they wear in the booth.
• ✅ Pets and cartoon characters work best from a sharp, front-on image.

• ❌ Group photos or a second face in the background — the model may pick the wrong person.
• ❌ Sunglasses, hands over the face, heavy shadows or strong filters.
• ❌ Tiny faces in a wide shot, or very low-resolution screenshots.

Only upload photos you have the right to use: yourself, friends who agreed, your pets, or characters you created.

The Hotel Lobby AI prompt

If you make the video directly with Seedance 2.5 (video edit mode, the original clip as @video1 with its audio removed, two photos as @image1 and @image2), this is the structure that held up in our tests. Describe the original performers by what they wear so the model knows exactly who to replace.

Edit the video: STRICT CHARACTER AND WARDROBE REPLACEMENT — TWO REFERENCES
@image1 — replacement for the LEFT performer.
@image2 — replacement for the RIGHT performer.

SOURCE PRIORITY
@video1 controls all movement, performance, lip-sync, facial expressions, gaze,
gestures, camera movement, framing, editing and timing. The reference images
control each replacement's face, hairstyle, skin tone, body, clothing and
accessories. Each reference image shows ONE person.

CHARACTER 1 — LEFT PERFORMER
Replace the performer on the left (describe their outfit) with the person in
@image1. Match @image1's face, hair, clothing and accessories exactly. Do not
keep the original performer's jewelry or sunglasses unless @image1 shows them.

CHARACTER 2 — RIGHT PERFORMER
Replace the performer on the right (describe their outfit) with the person in
@image2, with the same rules.

PERMANENT IDENTITY ASSIGNMENT
Left and right are the viewer's view at the start of @video1. Bind each
replacement to that performer for the whole clip. Never swap identities, blend
faces or exchange outfits.

EXACT SOURCE PERFORMANCE
Keep every gesture, mouth movement, head turn, camera move and cut at its
original time and speed. Same duration. No new poses, cuts or people.

ENVIRONMENT
Keep the burnt-orange booth, the hanging microphone and the lighting.
Do not copy the reference photos' backgrounds.

• Remove the audio from the source clip before uploading. In our tests, Seedance 2.5 rejected the edit when the clip still carried the copyrighted song.
• Add the original audio back afterwards, starting on the same frame as the clip. Because the motion and the audio start together, the mouth movements line up.
• Keep the left/right order of the photos the same as the order of the @image tags.

On CreateVision all of this — the clip, the prompt, the audio and the mixing — is done for you.

Three ways people make it, compared

Tutorials for this trend use three different techniques. They are not equally good at the thing that matters: keeping the original performance.

For Hotel Lobby, swapping the performers on the original clip gives the most faithful result: the booth, the camera and every gesture come from the real performance, and only the people change. That is how our tool works.

When it goes wrong: fixes from our test runs

Hotel Lobby AI video FAQ

Same idea, other trends: the On The Radar "679" AI video and the Raindance AI video.