Start here
How to make a fancy cafe AI video from two photos
The “first time in a fancy cafe” trend is a small, silent joke: someone sits on a round wall disc that looks like a stool, a stranger walks in and uses that disc as a table, and the first person has to climb down without saying a word. Fancy Cafe AI builds that 21-second scene from two photos. This guide is the exact workflow, from choosing the two photos to posting the finished MP4.
What you need before you start
- Two photos. One of the person seated first, one of the person who comes in later.
- An account. Sign in with Google so your credits and videos stay attached to you.
- 4 credits. Every 21-second video costs 4 credits. The One video pack is 4 credits, so one pack makes one video.
Step 1 — Decide who plays which role
The joke only lands if the two roles are clear. The person seated first is the one on the wall disc. The person who comes in later is the stranger who walks in, sits on the concrete ledge and puts the coffee cup on the disc.
You can put either person in either role. The generator keeps both faces locked to the photos, so the person you put in Photo 1 stays on the wall disc for the whole clip.
Step 2 — Choose two photos that match
Use one person per photo. A front-facing selfie or a clear upper-body photo works better than a group shot, a side profile or a photo where the face is partly hidden. Both photos should be sharp and evenly lit; a dark photo paired with a bright one makes the final grade fight itself.
Full checklist: Best photos for a fancy cafe AI video.
Step 3 — One format: 9:16
The scene is staged for a vertical frame: the two wall discs sit on the left and middle, so both people stay visible on a phone. The output is always 9:16, sized for TikTok, Instagram Reels and YouTube Shorts.
Step 4 — Start the render
Sign in, confirm the two photos, then press the generate button. The page charges 4 credits, uploads both photos to private storage and sends the scene to an image-to-video model. Most renders finish in 3–6 minutes.
You can close the tab. The render continues on the server, and the finished video appears in My videos when it is ready. If the model cannot finish, the 4 credits return to your balance automatically.
Step 5 — Download and post
The download is a clean MP4 with the cafe audio track, and no watermark. You can add a different sound inside TikTok, Reels or Shorts without fighting a second audio track.
For the cover frame, choose the moment the cup lands on the disc. That single frame explains the joke before the video starts.
What the 21 seconds contain
- The confidence. The seated visitor sips coffee on the wall disc like it is a normal bar stool.
- The entrance. The person who comes in later walks in and sits on the low concrete ledge.
- The cup. The newcomer places their coffee on the disc the first person is sitting on.
- The penny drops. The first person looks down, then at the newcomer, and freezes.
- The climb down. The visitor moves to the ledge and takes a tight-lipped sip.
Common mistakes
- Two people in one photo. The model cannot tell which face belongs to which role. Crop a copy around each person instead.
- A photo from the side. The model has less face to lock onto, so the likeness drifts.
- Heavy beauty filters. The generated face inherits the filter and stops looking like the person.
- Expecting a copy of the original TikTok. This is a new AI scene built from your photos, not a frame-for-frame copy of someone else’s video.
Frequently asked
Can I make a solo version?
Fancy Cafe AI is built around two roles, so it needs two photos. The other person can be anyone who agreed to appear.
Does it include the original sound?
No. The MP4 carries the cafe audio track. Add a different sound in the app where you post it.
How long does it take?
Usually 3–6 minutes. The job keeps running if you close the page.