The two people are on the wrong sides
Swap the input portraits before the next submission. Rename your source files left and right if you are making several combinations, and compare the previews against the intended cast.
THE COMPLETE WALKTHROUGH
How to Make a Hotel Lobby AI Video
Start with two separate portraits and a clear idea of who goes on each side. This guide takes you from choosing photos to checking the final performance, with practical fixes for awkward crops, mixed-up faces and inconsistent details.
Open the generatorBefore you start: choose the two photos
Use one person per image. A group photograph forces the model to decide which face matters and gives you less control over the cast. Choose a clear, front-facing portrait for each performer, with enough room around the head and shoulders to understand the person’s shape.
A waist-up photograph can give useful clothing information. Keep the eyes, jaw and hairline visible; heavy filters, oversized sunglasses, motion blur and a face hidden behind a hand all remove useful detail. You do not need a studio portrait, but a sharp everyday photo usually provides a better starting point than a tiny screenshot.
Try to match the two images in brightness. One dark nightclub photograph beside a strongly lit outdoor portrait asks the model to reconcile two very different faces and lighting conditions. If possible, photograph both people near the same window. Export the images as JPG, PNG or WebP, keeping each file within the upload limit shown on the tool.
1. Upload the left and right performers
Open the Hotel Lobby AI generator and add the first photo to Left performer. Add the second photo to Right performer. These positions are part of the creative setup, so choose them deliberately rather than treating the images as an unordered pair.
Look at the previews before continuing. Confirm that the intended person is visible in each slot, the face is not cropped at the chin, and neither image is the wrong file. Use the center swap button if the casting feels better the other way around. You can replace either image without starting the entire setup again.
For a friends video, consider who should lead the opening moment. For siblings or a couple, think about the contrast you want viewers to notice. Clothing, expression and height can all influence how the pair reads, even before any motion begins.
2. Choose the Hotel Lobby template
Select the original orange studio template for the recognizable performance format. Watch the reference clip so you understand its framing, microphone position and body movement. Your two photos supply the cast; the template supplies the scene and motion.
The template library also maps out friends, couple, pet-and-owner, costume and solo concepts. A concept card is a starting point for your casting, while the generator’s available controls determine what can actually be submitted. Choose the active template before expecting a particular duration, crop or replacement mode.
For the original workflow, you do not need a prompt. If you are using a prompt-enabled editing workflow, keep your direction short: identify the left subject, identify the right subject, then say what must stay unchanged. The prompt library includes copyable examples you can adapt to an editor that accepts reference images and video.
3. Generate and let the performance finish
Review the two portraits, the selected template and the credit amount beside the action button. Confirm that you have permission to use the photos and animate the people in them. Sign in if the generator asks, then submit the request once.
Generation involves more than placing two faces on one frame. The service has to keep their appearance consistent through movement, changes in pose and moments when one performer overlaps another. Waiting time depends on the provider and its queue. A 30-second video describes the planned clip length, not a promise that processing takes 30 seconds.
Use My videos to check the task status. Avoid repeatedly pressing Generate while a request is pending: a separate request can become a separate paid generation. If the service is still being configured, you can prepare the photos and explore the reference material while generation remains unavailable.
4. Review the result before downloading
Watch from beginning to end. Check whether each face belongs to the correct performer, whether hair and clothing remain stable, and whether hands or accessories distort around the microphone. Pause on the moments with the most movement instead of judging only the opening frame.
Listen separately for audio. The source track and the video image are distinct parts of the result; a silent browser preview may simply be muted, while a downloaded silent file needs a different fix. Check the player controls first. Do not assume that adding any popular track later gives you permission to use it in every context.
Download the finished clip when it is available. Provider-hosted result links may expire, so your task history should not be your only copy. Keep an original file, then make any cropped or captioned versions from that file rather than repeatedly compressing an already edited export.
Common problems and what to change
A face loses its likeness during movement
Try a sharper portrait with an unobstructed face and less extreme head angle. Remove heavy filters from the source and avoid a low-resolution crop pulled from a group photo.
Clothing changes between frames
Use a portrait that actually shows the desired clothing. Strong patterns, tiny jewelry and small printed logos are harder details to maintain than a simple, clearly visible outfit.
The vertical crop cuts someone off
Check the original framing before cropping. Keep both bodies within the portrait safe area. If that removes too much of the scene, use the original aspect ratio instead of forcing the crop.
A second attempt still looks wrong
Change one meaningful input at a time. Repeating exactly the same setup may produce a different result, but it does not fix a consistently unclear portrait. Review the credit cost before trying again.
How to do the Hotel Lobby trend for free
You can browse the example, choose your cast, prepare images and copy prompt ideas here without paying. Free personalized generation is a planned trial feature; the live generator shows the current sign-in and credit requirements. The free guide explains the distinction so you can plan without assuming an unlimited allowance.
If you use another service’s trial, check its current limits before uploading: credits, export quality, watermarks and access to video editing can differ. A free image tool is not necessarily a free two-character video replacement tool. Compare the actual output you can download rather than the word “free” alone.
Questions about how to make it
How long does a Hotel Lobby AI video take?
Processing time varies with the active model, source clip and queue. Check the task status in My videos. Clip duration and processing time are different, so a 30-second export should not be read as a 30-second turnaround promise.
Do I need editing experience?
No timeline editing is needed for the original template workflow. The useful decisions are choosing clear photos, assigning the correct sides and reviewing the completed clip. Cropping and captions are optional finishing steps.
Can I use two photos of the same person?
You can plan a double-cast concept with different portraits or outfits. Review the result carefully because the model may make the two versions look too similar. A dedicated solo mode is a separate planned option.
Does a failed generation use credits?
The current task flow returns credits when a generation is marked failed. A completed video that does not match your creative preference is different from a failed task; check the result before spending credits on another attempt.
YOUR TWO PHOTOS. YOUR MOMENT.