HOTEL LOBBY AI · TROUBLESHOOTING
Hotel Lobby AI Video Looks Wrong? Fixes for Blended Faces, Swapped Sides and More
Most bad Hotel Lobby AI videos come from the two input photos, not the tool. Blended faces, the wrong person on the wrong side and strangers appearing in the booth usually trace back to a crowded, dark or mismatched photo.
The short answer
These four fixes solve most cases:
- Use one person per photo. Crop group shots so each image shows a single face.
- Use bright, front-facing portraits. Both eyes visible, face large in the frame.
- Make the two photos look different. Different outfits, backgrounds or expressions help the model keep two identities apart.
- Check the slot order. In two-photo generators such as MigosAI, the first photo is the left performer and the second is the right.
The table below covers each symptom in detail.
Find your problem
| What you see | Likely cause | Fix |
|---|---|---|
| Both performers have the same face, or a mix of both | Photos are crowded, dark or too similar | Use two bright, single-person portraits that differ in outfit or background |
| A performer doesn't look like the person in the photo | The face is small, blurry, angled or covered | Use a sharp, front-facing photo where the face fills more of the frame; remove sunglasses and hats |
| The two people are on the wrong sides | Upload order | Put the left performer in the first slot and the right performer in the second |
| A third person or extra hands appear | Another person, arm or pet is partly visible in a photo | Crop so only one subject remains in each image |
| The booth isn't orange or the mic is missing | A different stage or tool setting was selected | In MigosAI, pick the Hotel Lobby stage; the other stages use different sets |
| The video is silent or uses a different song | Many tools export silent clips or an inspired-by soundtrack | Add the official "Hotel Lobby" sound from the TikTok, Reels or Shorts library when you post |
| The movement doesn't match the original choreography | The tool generates new movement instead of copying the source | Upload a motion reference clip, or use a motion-transfer tool if exact moves matter |
| The performance looks stiff or short | Very short duration or a low-detail photo | Try a longer duration and a waist-up photo that shows the hands |
| The upload is rejected | HEIC format, or a file over the size limit | Export as JPEG and keep each file under 10 MB |
| Progress stops at the same percentage | The job is between stages, such as photo or video review | Wait instead of resubmitting; check your saved videos later |
Every generation is a new attempt, so the same photos can give a better result on a second try. Switching to a higher resolution creates a new video; it does not upscale the one you already have.
The photo checklist that prevents most problems
Run through this before you spend credits. It takes a minute and avoids most failed generations.
- One subject per photo. No friends at the edge of the frame, no arm around a shoulder.
- Front-facing, eyes open. A slight angle is fine; a profile is not.
- Face large and in focus. A waist-up shot works well, as long as the face is still easy to recognize.
- Bright, even light. Avoid backlit windows, harsh shadows and night photos.
- Nothing covering the face. Remove sunglasses, masks, wide hats and heavy filters.
- Two photos that look different. Similar lighting is fine, but different clothes or backgrounds help keep identities apart.
- The right format. JPG, PNG or WebP under 10 MB. Export iPhone HEIC photos as JPEG; renaming the file does not convert it.
- Permission to use both faces. Only upload photos you own or have consent to animate.
Want the same person in both roles? Use two different photos of them, ideally in different outfits. Making a pet version? See the cat and dog photo guide.
Regenerate, change the photos, or change the workflow?
Regenerate when the result is close but one moment looks off, such as a strange hand or an odd expression. Results vary from run to run.
Change the photos when the same problem shows up twice. Blended faces, wrong identities and extra people rarely fix themselves on a retry with the same inputs.
Change the workflow when the tool can't give you what you want. Hotel Lobby AI tools work in three different ways:
| Workflow | How it works | Where it struggles |
|---|---|---|
| Two-photo template (for example, MigosAI) | Generates a new orange-booth performance from two photos; no prompt needed | Movement is newly generated, so it won't copy the original moves frame for frame |
| Prompt-based generator (for example, Dreamina or Media.io) | You upload photos and write the scene in a prompt | Vague prompts cause side swaps; spell out left and right, and ask for no extra people or face blending |
| Motion transfer or face swap | Swaps your subjects into an existing performance clip | Needs a source video and more setup; reusing footage raises rights questions |
On MigosAI (migosai.design), test new photos with a 5-second, 480p Wan template clip, which costs 15 credits. Generate the longer, sharper version only once the pairing looks right. See the credits guide for the full cost table.
Frequently asked questions
Why do the two faces blend together in my Hotel Lobby AI video?
Face blending usually happens when the reference photos are crowded, poorly lit or too similar. Use two separate, bright portraits that look clearly different.
Why is my friend on the wrong side?
Most two-photo tools place the first upload on the left and the second on the right. Swap the uploads and generate again.
Why doesn't the AI video look like me?
The face in your photo is probably small, angled, blurry or partly covered. A sharp, front-facing photo with your face large in the frame works best.
Why is my Hotel Lobby AI video silent?
Many tools export a silent clip or a soundtrack inspired by the trend. Add the official "Hotel Lobby" sound from your app's music library when you post.
Can I fix a finished video without paying again?
Usually not. Each fix is a new generation, which uses credits again. Testing new photos with a short, low-resolution clip keeps the cost down.
Why are the moves different from the original Quavo and Takeoff performance?
Template generators create new movement inspired by the scene rather than copying the source. If the exact choreography matters, use a motion reference clip or a motion-transfer tool.
Sources
Create your Hotel Lobby AI video
Two photos. A built-in duo template. No prompt to write.
Open the Hotel Lobby AI Video GeneratorSign-in and credits required. Check the cost before generating.