How to Make the Hotel Lobby AI Video with Higgsfield

The Hotel Lobby AI trend — two performers trading verses in an orange studio booth — was largely popularized by creators using Higgsfield. This tutorial walks through the exact Higgsfield workflow people use, including the prompt, and then shows a faster route if you would rather skip the setup.

Hotel Lobby AI is an independent tool and is not affiliated with Higgsfield. This is an unofficial tutorial based on publicly shared workflows.

What you need

• A Higgsfield account (higgsfield.ai) with enough credits for a video generation.
• The source video — the orange-booth performance clip the trend is based on. Tutorial authors typically share it in their post; only use footage you have the right to reuse.
• Two photos, one for each performer you want to cast: one subject per photo, front-facing, evenly lit.

Step 1: Open Higgsfield Genjutsu

Log in, hover over Video in the navigation, and open Higgsfield Genjutsu — the motion-transfer tool that swaps the people in a source video while keeping the original performance, camera, and sound.

Step 2: Upload the source video and your two photos

Add the source video first, then your two character photos. Pay attention to the attachment labels Higgsfield assigns (for example @video, @image_1, @image_2) — your prompt has to reference exactly these labels, or the swap targets the wrong person.

The photo order matters: the first photo replaces the performer who starts on the left of the source video, the second photo the one on the right.

Step 3: Paste the prompt

The most widely shared prompt for this trend comes from creator Nate Herkelman's walkthrough:

Replace the man on the left side of the source video with the person in @image_1, and replace the man on the right with the person in @image_2. The original sound and movements should remain the same. Just swapping characters.

Adapt the wording to your cast (two women, a person and a pet, and so on), and keep the @ tags matched to your actual attachment labels.

Step 4: Check settings and generate

Select Genjutsu as the model, pick a quality tier, and confirm the credit cost shown on the page. Then generate and review the result — check that both identities survive head turns and that the orange set and hanging microphone stay intact.

Tips from creators who got clean results

• Match @ tags to attachment labels, not to the order you think you uploaded in. Mismatched tags are the top cause of swapped or blended faces.
• One subject per photo. A group shot gives the model an ambiguous read.
• Test short before long if the tool lets you trim the clip — verify identity on a few seconds before spending credits on the full take.
• Audio: motion transfer keeps the source sound; when you post, make sure the track is cleared through the platform's licensed sound library.

A simpler route: two photos, no prompt

The Higgsfield workflow is powerful, but it asks a lot: find the source video, manage attachment labels, write a prompt, tune settings, spend credits per attempt.

Our Hotel Lobby AI video generator strips that down to the two things that actually personalize the video — your two photos. The orange booth, the hanging microphone, the camera framing, and the left/right casting are built into the template, so:

| | Higgsfield Genjutsu | Hotel Lobby AI | | ------------- | ---------------------------- | --------------- | | Source video | You find and upload it | Built in | | Prompt | You write and match @ tags | None | | Inputs | Source video + 2 photos | 2 photos | | Setup choices | Model, quality, credit cost | Generate button |

If you enjoy tuning workflows, Higgsfield gives you more dials. If you just want your duo in the booth, upload two photos and generate — the result is a vertical 9:16 clip ready for TikTok, Reels, or Shorts.

FAQ

Is this the official Higgsfield tutorial? No. It is an independent guide based on workflows creators have shared publicly. Higgsfield's interface changes often, so trust what the current page shows over any tutorial.

Can I use the same prompt in other tools? The prompt structure — left/right identity binding plus "keep sound and movements" — transfers to other motion-transfer tools, but attachment label syntax differs per platform.

Does the Higgsfield version include the song? Motion transfer preserves the source video's audio. Whichever tool you use, add the track through TikTok or CapCut's licensed library when posting so the audio is cleared.

Make your Hotel Lobby AI video →