Make an original two-character performance in an orange studio. The Hotel Lobby AI trend recasts the duo in a spare performance setting: two distinct characters, one hanging microphone, and contrasting gestures.
The prompts here are free. External image and video generation may cost money. MuseShuffle does not generate videos or accept photo uploads.
This is an independent MuseShuffle guide, with no affiliation to the artists or COLORS. The templates below use your own characters and staging; they do not include the original recording, lyrics, or performance footage.
Choose the right route
A. Create an original orange-studio duo
Character references → composed still → image-to-video. Prepare references for two distinct characters, create a shared scene, then animate that first frame in a tool that accepts image input.
Use the templates below for composition and simple movement. This route does not guarantee the original performance's movements, rhythm, or lip sync. A song title in a prompt is not a timing track.
B. Follow a reference performance
Use a documented performance-transfer workflow when acting or dialogue timing matters. Runway's official multi-character Act-Two tutorial uses separate driving performances for each character, individual generations, and compositing in a video editor. Act-Two takes single-character inputs; this is not a one-pass two-person transfer.
Its shared-video workflow isolates each character and combines the results over a base scene. For gesture transfer, the tutorial instead describes separate character images and compositing. Neither workflow is a verified recreation of the Hotel Lobby performance.
Record your own performance or use footage you have permission to adapt. Follow the tool's input instructions; do not feed the full-body duo template into a single-character workflow unchanged.
Example composition for route A: two distinct characters, one suspended microphone, flat orange studio background.
Copy the prompt templates
Starting templates. Output varies by tool and model version — adapt the wording to the tool you use, and attach your references through that tool's own controls.
For route A, try a 16:9 composition and a short 5-second motion test if supported. These are adjustable suggestions, not settings available in every model. For vertical output, recompose both characters with room around them instead of cropping a wide shot.
01 / Compose the first frame
Image prompt
Create an original studio performance still with exactly two distinct characters based on my own or licensed reference material. Character A is [appearance, hairstyle, and outfit], standing on the viewer's LEFT. Character B is [different appearance, hairstyle, and outfit], standing on the viewer's RIGHT. Keep their identities, faces, outfits, and bodies separate. Show both full bodies from head to shoes, with clear space between their silhouettes and generous margins above and below. Use a seamless orange studio wall and floor, soft even lighting, and one suspended microphone centered between them, hanging from the top of the frame. Both face the camera in relaxed, distinct poses, with visible hands resting naturally. Eye-level wide composition, balanced spacing, clean background, no text or logos.
02 / Animate the approved frame
Video prompt
Animate the supplied two-character studio image as a single continuous shot. Keep the camera fixed at eye level with no pan, zoom, or cuts. Character A on the viewer's LEFT makes a small shoulder bounce and one slow open-hand gesture near the waist. Character B on the viewer's RIGHT gently nods and shifts weight once, keeping both hands relaxed. Let their actions differ rather than move in unison. Preserve each character's face, hairstyle, outfit, body proportions, and left/right position throughout. Keep both complete bodies and their hands within the frame. Maintain the orange studio, lighting, and the single suspended microphone in place. Use restrained, natural movement with clear separation between the characters.
The video template directs movement only. It does not synchronize speech or recreate choreography. If exact performance timing matters, use route B and review its limitations first.
How to make the Hotel Lobby AI video, step by step
Prepare your material. Choose two original characters, or people and images you have permission to use. Write a distinct description for each. One clear reference per character can be useful, but the required number and framing depend on the tool; two photos and full-body photos are not universal requirements.
Generate and inspect the first frame. For route A, assign A to the viewer's left and B to the right. Check separate identities, hands, space around the bodies, and one microphone before spending credits on motion. Fix the still before animating it.
Generate movement. Use the approved still as the image-to-video input and start with restrained actions. For route B, follow the linked multi-character workflow, preparing each performance separately and compositing as required.
Check the whole clip. Watch at normal speed and pause at several points. Look for changing faces or clothes, merged limbs, side swaps, and clipped feet. Revise one issue at a time; a good first frame does not guarantee stable later frames.
Add audio you may use. Add your own recording or appropriately licensed audio in your editor. Check the terms for your intended platform and use. Naming or crediting a song does not supply an audio license. Adding music afterward also does not make existing mouth movements match it.
Is there a free way to make a Hotel Lobby AI video?
Yes — but "free" means something different at each step. Composing the still frame is the
cheapest part and is usually doable on a free tier; animating it is where watermarks, short clip
lengths, and export limits appear. Three routes, cheapest first.
1. Use a free entry tier for the still frame
Most image generators hand new accounts a one-time or daily allowance. Spend it on the
first frame, not on motion: the still is the step you will redo most often, and a
bad first frame wastes every credit spent animating it.
Check five things before you generate: whether the feature you need is included in the free
tier, the credit cost per attempt, the output resolution, whether exports carry a watermark, and
the usage terms. Expect several attempts per usable frame — a clean two-character composition
rarely lands on the first try.
2. Animate with a ready-made template or a short free render
Editing apps ship templates that take two photos and return a short clip. This is the
lowest-effort way to see the trend on your own characters, and in most free versions it costs
nothing but a watermark. Keep the clip short — a few seconds — and export at the highest setting
the free tier allows.
If you animate with an image-to-video model instead, feed it the still you already approved and
start with a 5-second test rather than a long render. Free allowances are usually measured in
seconds, not minutes.
3. Skip generation entirely (fully free)
Nothing about this trend requires a generated video. A single still — two characters, a flat
orange backdrop, one micro phone hanging on a thin cable — carries the joke, and you can build it
from photos you already have with a free image editor: cut both subjects out, place them on a flat
orange background, add a slim line for the cable.
This route has no credits, no watermark and no waiting. It is also the safest one: your own
material, your own editing, nothing generated.
What free tiers generally will not give you: a watermark-free HD export, long durations, or
performance matching. Treat every free allowance as a limited number of attempts, and re-check the
plan page before you pay for anything — limits change without notice, so verify them at the
tool itself.
If you need performance matching (route B)
For route B, Runway's
Act-Two documentation currently lists Standard-plan access or higher and credit-based
generation. It also distinguishes gesture control with character images from facial animation with
character videos. Check that page for current requirements before paying. Documentation checked
September 27, 2026.
Common problems and useful fixes
Characters blend into one another
Make their silhouettes, outfits, and descriptions clearly different. Leave space between them in the still and reduce overlapping gestures. Where supported, assign each reference to a specific subject. Regenerate a mixed first frame before animating.
Left and right switch
Use viewer-relative directions consistently. Keep the same first frame and avoid crossing or camera movement. If your tool offers subject assignment, check it; repeating left/right text alone cannot guarantee tracking.
Hands warp or grow extra fingers
Inspect the still, simplify the gesture, and keep hands apart from faces and the microphone. Try a shorter segment or less motion. Reject damaged frames rather than relying on a longer prompt to repair them.
Heads or feet are cropped
Recompose with wider margins and the intended output ratio. The full-body framing here is a route A design choice. Performance-transfer tools may require closer framing: Runway's linked guidance recommends single-subject, waist-up-or-closer inputs.
Movement or lip sync does not match
Route A is an original animation, not performance matching. For controlled acting, use an appropriate reference-performance workflow. Align the intended audio and performance before generation where required, and check both characters separately. No text template guarantees the original choreography or delivery.
Last updated . AI video tools change often — check the linked documentation for current limits.
Frequently asked questions
What is a Hotel Lobby AI video?
In this context, it is an AI reinterpretation of the orange-studio duo performance associated with Quavo and Takeoff's HOTEL LOBBY on A COLORS SHOW. This guide helps you plan an original scene or choose a reference-performance workflow.
Which photos do I need?
Use clear material you own or have permission to use for each character. Separate references can help distinguish the pair, but each tool sets its own number, format, and framing requirements. Full-body source photos are not always required even when the desired composition shows full bodies.
Are these prompts free?
Yes — reading, copying and adapting both templates is free and needs no account. For the generation steps, see the three free routes: a free tier can usually cover the still frame, while watermark-free HD export and performance matching normally require a paid plan.
Can MuseShuffle generate the video directly?
No. This page provides text templates and instructions. It has no video-generation service, photo uploader, or connection to a generation API.
Do I need a reference video?
Not for route A's original image-to-video scene. Route B uses driving performance footage, with separate inputs and editing steps depending on the tool. The documented Runway workflow processes characters individually; use your own or permitted footage.
How should I handle the audio?
Use your own or licensed audio and check that its terms cover your use. For lip sync, plan audio alongside the performance workflow. Putting a song under an unrelated animation will not synchronize its mouths or gestures.
What if the characters do not stay consistent?
Start with a clean, clearly separated first frame, distinctive clothes, simple actions, and a short test. Use supported reference controls, review the entire result, and regenerate when identity changes. These templates cannot guarantee consistency.
