Quick answer
Starrd's I Thought We Were Having A Nice Date template recreates the viral restaurant-date scene from two photos. Photo 1 is the star, who flirts "No, no, no…", laughs, stands up and says "I thought we were having a nice date", then breaks down crying. Photo 2 is the date, who asks "What?" and pleads "We are, we are, sit down." The full 21-second scene is rebuilt shot for shot, all nine shots cut on the original's timing, with the original dialogue and real lip sync, in a dim old-fashioned restaurant booth. Works with people and pets. Vertical 9:16, Seedance 2.5, 200 credits (about $10).
Two photos in, and out comes you and your date recreating the "I thought we were having a nice date" scene: one of you laughs, stands up and breaks down crying in a restaurant booth while the other begs them to sit down. All 21 seconds, lip-synced to the original dialogue. Here it is with two chimps on a very fancy date:
I Thought We Were Having A Nice Date
Two photos. One of you laughs, stands up and breaks down crying; the other begs them to sit down. The full 21-second scene, lip-synced.
What You Get
- The whole scene. All 21 seconds and all nine shots, cut on the same frames as the original.
- Two roles that never swap. Photo 1 is the star who breaks down, photo 2 is the date.
- Real lip sync. Each of you mouths only your own lines; when the other one talks off screen, your mouth stays shut.
- The original camera. Tight long-lens close-ups, the wide shot of the stand-up, the high angle down at the date and the low angle up at the star, with the two of you always looking at each other, never at the lens.
- The whole arc. A flirty smirk, a giggle, a too-bright smile, a hurt little smile, then the ugly cry and a teary "It's okay".
- People or pets. Cats and dogs cry with their own faces; chimps stand up and reach like the actors.
- 21 seconds, vertical 9:16, ready for TikTok, Reels and Shorts.
What the Meme Actually Is
It's a scene from the trailer for Obsession (2026), posted on TikTok by @universalpics. A couple sits in a dark wood booth at a restaurant. She teases him, "No, no, no…", and laughs. He asks "What?" She stands up, still smiling: "I thought we were having a nice date." He reaches out, "We are, we are, sit down", and her smile caves in: "Why does it matter? I thought we were having a nice date." He keeps saying "We are, we are, we are" as she sits back down, crying, and says "It's okay."
Why it suits AI: the joke is the swing from flirty to falling apart in four seconds. Two photos let you cast your partner, your best friend or your cat as the one who loses it.
The Scene, Shot by Shot
| Time | Shot | What happens |
|---|---|---|
| 0:00 | Star close-up | Flirty and giggly: "No, no, no…" |
| 5.8s | Date, medium | Baffled: "What?" |
| 8.2s | Star close-up | Laughing: "No, no, no, don't do that!" |
| 9.7s | Wide | The star stands up at the table: "I thought we were having…" |
| 11.2s | Star, over the shoulder | A too-bright smile: "…a nice date." |
| 12.5s | Date, high angle | Reaching out: "We are, we are, sit down." |
| 14.3s | Star, low angle | "Why does it matter? I thought we were having a nice date", and the crying starts |
| 17.6s | Date, medium | "We are, we are, we are." |
| 19.4s | Star close-up | Sitting back down in tears: "It's okay." |
Who plays who: photo 1 cries, photo 2 pleads. Put the person who'd be funniest falling apart first.
Your Look
- Your own face, hair and clothes. Nobody gets the original actors' outfits.
- Your own glasses and hats. If your photo has sunglasses or a cap, you keep them; if it doesn't, you don't get any.
- Costumes stay on. Wigs, veils, monocles, chains: whatever is in the photo is in every shot.
Use clear, well-lit face photos. Most shots are close-ups, so faces carry the video. Sunglasses hide some of the tears, so a photo without them gives the biggest breakdown.
The Fastest Way
I Thought We Were Having A Nice Date
Two photos, the full 21-second scene with the original dialogue, lip-synced.
Two photos, 200 credits (about $10), no prompt. It renders on Seedance 2.5 at 480p, vertical 9:16.
Build It Yourself
1. Scrub the burned-in subtitles. Their letters show up as raised shapes in a depth map. Find the white caption pixels in each frame of the trailer and inpaint them first.
2. Turn it into a depth map, one shot at a time. Run it through a depth-estimation model (Depth Anything V2 works) and set the contrast per shot. Keep the close-ups sharp so the jaw movement survives; that's what the lip sync follows.
3. Rebuild it on the model's own frame grid. If the model renders at 24fps, resample the depth video to 24fps with every cut on the same frame number as the original, so nothing gets stretched.
4. Put only the on-screen voice inside the depth video. Mute each line in the driver whenever the speaker is off screen. Otherwise the model moves the wrong mouth, and can even pull the wrong person into the next shot. Lay the full original dialogue back on at the end.
5. Make one character sheet of the two of you. A wide shot of both in the booth, then a face, three-quarter, profile and full figure of each, with the eyes looking off to the side, never at the lens.
HARD CUT at exactly 00:14.25 to a close-up of THE STAR from a slightly low angle. Only THE STAR is on screen, from the first frame to the last. Eyes down toward the right edge of the frame, at THE DATE: a hurt, confused little smile on "Why does it matter?", then the face crumples on "I thought we were having a nice date": brows pulled up hard in the middle, chin wobbling, eyes flooding, mouth stretched open in a sob.
6. Drive it with the depth map and the sound. Feed the sheet, both photos and the depth video to a model that copies motion from a reference video (we recommend Seedance 2.5). Then write the four things a depth map can't carry:
- A screen timeline: who is visible, by absolute time, with a "HARD CUT at exactly" line for every cut.
- Eyelines and lenses: who looks toward which edge of the frame, and the framing and camera height of each shot.
- Lip sync: each mouth follows the voice syllable by syllable and stays shut in silence.
- The tears: say exactly how ugly the cry is, or the model gives you a polite, dry-eyed sniffle.
Common Mistakes
Leaving the off-screen lines in the driver. Her "Why does it matter?" starts while he's still on screen. With it in the driver audio, the model held his face for two extra seconds and her breakdown started late.
Blurring the faces in the depth map. It hides the jaw, and the mouths stop following the words.
Forgetting the eyelines. Without them, close-ups (and pets especially) stare straight into the lens instead of at each other.
Giving apes the pet rules. Cats and dogs need real paws and four legs. Chimps don't: let them stand up and use their hands, or they come out contorted.
FAQ
How many photos? Exactly two: the star and the date.
Does it work with pets? Yes: cats, dogs and chimps.
How long is it? 21 seconds, vertical 9:16.
What does it cost? 200 credits, about $10.
Related Reading
- How to Make the 'Deep Down In My Heart… I Still Love You' AI Crying Meme: another two-person crying meme, lip-synced to the original audio
- How to Make the 'Like 700 Mile Per Hour' AI Interview Meme: a two-person interview meme, lip-synced line for line
- How to Make the 'Hit the Road Dude' AI Video: another viral clip, lip-synced line for line
- Viral AI Video Trends 2026: every trend we've turned into a template this year
