Hotel Lobby AI examples
Read this first: every example below was made from AI-generated portraits of fictional adults, characters and pets, not from photos of real people. Each clip came out of the production generator and was not retouched or edited; the files may be re-compressed for the web. The thumbnails under "Made from" are the exact images that were uploaded.
Settings used for every clip
- Output: vertical 9:16, 768×1344, about 10 seconds, no watermark. 4:3 is also available.
- Model: MiniMax H3 Max, run through fal.ai, with the built-in orange-booth scene. No prompt was written.
- Lines: the preset lines for the voice style, shown under each clip. They are original lines, not lyrics from any song.
- Generation time: usually a few minutes depending on demand. Exact times were not recorded for these clips.
The clips

“Yo, I walk in, lights come on, everybody knows my song…”“Hold up, now it's my turn, bring the beat, watch me burn…”

“Stay with me tonight, under the golden light.”“I will sing it back to you, every word is true.”

“Good morning everyone. I am very, very serious.”“Me too. Very serious. Watch my serious face.”

“Good morning everyone. I am very, very serious.”“Me too. Very serious. Watch my serious face.”

“Yo, I walk in, lights come on, everybody knows my song…”“Hold up, now it's my turn, bring the beat, watch me burn…”

“Psst. Did you drink water today? Good. I'm proud of you.”“Now take a deep breath, bro. Tomorrow will be better.”

“Oh great, another Monday. I just love working hard for no thanks.”“Same. I'm exhausted, and nobody ever says thank you.”
The second source portrait (a pug) was not archived, so it is not shown. “Welcome back to the show. Today's topic: why does the human leave every morning?”“Great question. I think they love the mailman more than us.”
What these examples do not show
- They do not show results from real photos of real people. How well a face is kept depends on the photo: front-facing, well-lit, one subject, at least 512px on the short side works best. Side profiles, sunglasses, two people in one photo and low resolution give worse results.
- Very tall or very wide images are rejected before upload. Photos with children, public figures or unsuitable content are blocked by automated checks before generation, and a blocked photo does not use up a video.
- If a video fails to generate, it is returned to your balance automatically.
What a video includes
See what a video includes and does not include, including the full preset lines for all six voice styles. Ready to try it? See prices: one video is $7.99.