How to Make a Hotel Lobby AI Video with Two Photos

Start with two separate portraits

Choose one photograph containing only the first person and another containing only the second. A group photo asks the system to interpret several faces before it can establish the intended pair. Separate portraits make the assignment explicit. Use photographs you have permission to upload and discuss the intended result with the other person, particularly if you plan to post it publicly.

Favor a face that is large enough to inspect, evenly lit and unobstructed. Hair across the eyes, a hand over the mouth or a very small face in a distant photograph removes useful information. You do not need a professional headshot. A clear phone portrait generally gives you a more useful reference than a heavily filtered image whose features have already been changed.

Keep both original files somewhere accessible. The browser prepares supported JPEG, PNG and WebP images for upload, so reducing a large photograph manually is not normally necessary. If a phone format is rejected, save an ordinary image copy first. Check that the copy still shows the whole face before selecting it.

Check the left and right performers

Upload the first and second photographs into their respective slots. Look at the thumbnails rather than relying on filenames: phone libraries often contain several similar exports. The left and right assignment determines which person is mapped to each role. Use the swap control if you want the opposite pairing before you submit the job.

This choice is especially useful when one person wants a particular part of the performance. A portrait does not contain instructions about who should lead or react. The chosen source clip supplies that sequence. Watch its preview, identify which position performs the moment you like, and place the corresponding portrait in that slot.

Choose the source clip and frame before quality

The three template names are Classic, Verse and Encore. They are intended to represent different performance excerpts, each with portrait and landscape framing. Judge them by the playable source preview when available. A name alone cannot tell you whether the gestures, timing or opening pose suit the two people in your photographs.

Choose 9:16 for a tall composition or 16:9 for a wide one. Think about where the clip will be watched and how much room both performers need. Cropping a finished landscape video into a narrow frame can remove one person. Choosing the intended frame at generation gives you a better starting point for a later edit.

Every Hotel Lobby AI video on this site lasts 15 seconds. Changing the frame or quality does not unlock a longer song, a new soundtrack or an editable timeline. If you need those controls, decide before spending credits whether a separate video editor or a different creation workflow fits the project.

Budget your Hotel Lobby AI video

Standard is the default and uses 300 credits. Pro uses 400. Pro is an alternative processing tier, not a promise that every face, hand or syllable will be perfect. Start by fixing weak input photographs rather than assuming a more expensive setting can reconstruct detail the reference never provided.

The one-time packs are $9.90 for 450 credits, $19.90 for 1,250 and $39.90 for 3,150. Credits do not expire, and these packs do not renew automatically. The smallest pack covers one Standard generation with 150 credits remaining, or one Pro generation with 50 remaining. It does not cover two full generations at either setting.

A fresh generation is a new credit decision. Changing a portrait after seeing a completed result does not revise the previous video for free. Before submitting, check both thumbnails, positions, template, frame and quality together. This short review is more useful than discovering a mistaken assignment after the task has already been accepted.

Generate once, including through login or payment

Press Generate when you are ready to create the video. If you need to sign in, complete the authentication flow. If your balance is insufficient, choose a credit pack and finish checkout. The site is designed to resume that same explicit generation request after the required step, so returning from payment should not require a second Generate click.

Simply signing in, restoring a draft or purchasing credits does not submit a job by itself. The distinction matters when you are only preparing photographs for later. Once a job is accepted, look for its task status before pressing anything again. Repeatedly submitting because a result is not instant can create additional work when you actually meant to check the first task.

A browser draft helps preserve your photographs and settings during this process, but it is not a permanent backup. Drafts are kept locally for up to 24 hours. Changing browsers or clearing browser storage can remove that local state. Keep the original portraits until you have retrieved the finished video.

Follow the task, then inspect the whole result

After acceptance, processing continues in the background. You can return to My creations to check your Hotel Lobby AI video; keeping the generator visible is not what makes the backend continue. Completion time varies, so use the status shown for the job rather than a countdown from an unrelated tutorial.

If a task appears unresolved, keep its identifier and consult help before starting a replacement. A confirmed generation failure returns the charged credits once. An ambiguous provider response needs reconciliation because the original request might still finish. A slow job is not automatically a failed job, and a new submission does not repair its status.

Play the completed clip from beginning to end with sound. Check that each person stays on the intended side, faces remain recognizable during movement, and the mouth movements remain plausible with the soundtrack. Pause where a performer turns or a hand crosses a face. A good opening frame does not prove that the following fourteen seconds are usable.

Download a copy and choose how to share it

Download the finished file and play that local copy before uploading it elsewhere. This separates a generation problem from a browser playback or social-platform processing problem. If the local file has sound but a post does not, investigate the destination's audio handling rather than paying to generate the same performance again.

On a phone, the browser's download location and the photo library may be different places. After downloading, check the browser's downloads list or your device's file manager. Open the file from there before deciding that saving failed. If you want it in your photo library, use the device's available save or share controls; the exact labels depend on the operating system and browser.

For a simple final check, send yourself the file through the same route your audience will use and open that copy. Confirm that both people remain visible and the full fifteen seconds play. If a messaging app compressed the video heavily, compare it with the original download before changing any generation settings. The original result and a forwarded copy may have different visual quality.

Your creations are private to your account unless you choose to share them. A share link gives access through a revocable token; treat that link as something recipients can forward. Downloading a file gives the recipient a separate copy that cannot be recalled by later revoking the original link.

Completed outputs have no automatic expiry under the current product rules, but keeping your own copy is still sensible. Delete unwanted videos through the account controls when appropriate. When posting a convincing remake, describe it as an AI recreation and identify the performance reference so viewers can distinguish the joke from footage of a real event.