The prompt
Paste this into an image-to-video model along with a photo of the two people, or one photo per person if the tool takes reference images:
Two friends perform a rap together inside a small music-video booth with
seamless bright orange walls and floor, lit by warm soft studio light.
Medium close-up, vertical 9:16. They stand side by side facing the camera,
trading lines: one raps with hand gestures while the other nods to the beat
and hypes, then they switch. Confident, playful energy, big grins, shoulder
bounces. Handheld camera with slight sway, quick push-in on each verse.
Keep both faces identical to the reference photos. No text, no logos.What each part does
| Part | Why it's there |
|---|---|
| "seamless bright orange walls and floor" | The booth colour is what makes viewers read it as the Hotel Lobby trend |
| "medium close-up, vertical 9:16" | The trend is shot close on the faces and posted vertical |
| "trading lines … then they switch" | The call-and-response is the point of the clip; without it you get two people dancing |
| "handheld camera … push-in" | A locked-off camera looks like a stock clip |
| "Keep both faces identical to the reference photos" | Pushes the model to keep your faces instead of inventing new ones |
| "No text, no logos" | Models like to add captions and watermarks |
Why prompt-only results drift
- No song. A video model makes visuals. You still need the audio, and the original recording is copyrighted.
- No real lip-sync. Mouths move, but not to the words. Lining them up takes a separate lip-sync tool.
- Faces wander. Over 10 to 15 seconds of movement, two faces tend to blend or change. The more motion, the more drift.
- Cuts and timing. The trend has its own rhythm. A prompt gets one continuous shot; matching the cuts means generating several clips and editing them together.
That is why most prompt workflows end up as several renders plus an editor. It's a fine way to experiment, just not a quick way to post the trend.
The no-prompt way
Hotel Lobby AI on RapVideo AI has the booth, shots, timing and audio built into the template:
- Choose Two photos (or One photo of both).
- Add a clear, front-facing photo of each person.
- Optionally type a topic for an original rap: names, an occasion, an inside joke.
- Tap make. The video is ready in about 2 to 3 minutes, lip-synced.
A standard 12-second 720p video is 10 credits, and a failed render is never charged.
Questions
Which AI tool is the Hotel Lobby AI made with? People use different tools. General video models work with a prompt like the one above plus editing; template tools like RapVideo AI make it from photos. See Hotel Lobby AI apps compared.
Can I write my own lyrics into the prompt? A video model won't sing them. On RapVideo AI the topic you type becomes the lyrics of an original rap, performed by the two of you.
Why do the faces not look like us? Usually the photos. See Hotel Lobby AI doesn't look like me.