"I thought we were having a nice date" is the line from the restaurant scene in the 2026 horror film Obsession, used as a meme for a smile that turns into laughter and then tears. You can make your own 15-second version of it from one photo on the I thought we were having a nice date AI page, with the film's original dialogue on it.
Key takeaways
- The scene is a smile, a laugh and a collapse, which is why any small disaster fits the caption.
- One photo plays the woman; a second adult photo, with their consent, plays the date. The two never touch.
- The audio is the film's original dialogue, so the voice is the actress's, not yours.
- A standard 15-second 720p video is 13 credits, and a failed render returns them.
- Never upload photos of children, celebrities or anyone who has not agreed.
In this guide
- Why did this one restaurant scene become a meme?
- How do I write a caption that works with it?
- What is the difference between the single and the two-person version?
- Why do the two people never touch?
- Which photos give the best result?
- Whose photo can I upload?
- Why is the voice in the finished video the original actress and not me?
- Should I label it as AI when I post it?
- Can I make this with ChatGPT, Sora or Veo instead?
- Where does the line come from, and what do reports say about the film?
- What does it cost, and what if it goes wrong?
Why did this one restaurant scene become a meme?
It packs a full mood swing into a few seconds, and everyone has lived a small version of it. In the scene from Obsession, a young woman in a dim booth smiles and says "No, no, no", stands up laughing with "don't do that!", and then, as her date asks her to sit down, breaks into tears over the line "I thought we were having a nice date."
The structure is what makes it easy to reuse. There is a smile, a laugh and a collapse, in that order, so any petty disaster fits: a cancelled plan, a bad text, a spilled drink. The scene also needs no setup, so a viewer who has never seen the film still reads it at once.
How do I write a caption that works with it?
Name the small thing that ruined the evening and keep it short. The joke is the gap between a calm start and a sudden collapse, so the caption should describe the calm part: "Dinner was going great" or "Day three of the diet". The video supplies the collapse.
Avoid explaining the scene. Viewers who know it will laugh at once, and viewers who do not will follow the face. A caption in the first person works best, because the person on screen is you.
What is the difference between the single and the two-person version?
With one photo, you play the woman who smiles, stands up and cries, and an ordinary adult stranger sits across the table as your date. With two photos, the second person takes the date's seat, so the cutaways to the other side of the table show someone you chose.
Your photo takes her part whatever your gender. On the I thought we were having a nice date AI page the second photo is optional, and the price is the same either way: 13 credits for a standard 15-second 720p video.
Why do the two people never touch?
Because the original scene is built on distance. Both people sit at the same table, one stands beside it laughing while the other stays seated and gestures for them to sit back down, and nobody reaches across. The videosi version keeps that rule: no hug, no kiss, no held hands.
It is also a safety choice. Keeping the two apart makes the result easier to share with a friend or partner, and it keeps every render a simple two-adult restaurant scene.
Which photos give the best result?
Pick one sharp, front-facing photo with your face large in the frame. Most of the 15 seconds are close-ups, and the laugh and the tears move the face a lot, so a small face, sunglasses or a heavy filter gives the model little to hold on to.
- Use even light and a plain background.
- Show your own hair and clothes as you want them; nothing like a wig or a sweater is added for you.
- For the second photo, use a similar clear shot of one adult, not a group.
Likeness is not guaranteed. If a render fails, times out or is blocked, its credits come back automatically.
Whose photo can I upload?
Yours, or an adult friend or partner who has said yes. Do not upload photos of children, celebrities or anyone who has not agreed. The rule is simple and it applies to the second photo as much as the first: if you would not ask them before posting, do not upload them.
Why is the voice in the finished video the original actress and not me?
The 15 seconds come with the film's own dialogue, laid over the video: "No, no, no", "don't do that!" and the title line. That means the voice you hear is the actress's, not yours, and your mouth moves loosely rather than matching each word exactly.
This is why the clip feels like the scene rather than a dub. If you want your own voice, add it in an editor after you download. Also note that a platform may mute or limit the clip's audio when you post it.
Should I label it as AI when I post it?
Yes, say so. A short "made with AI" in the caption costs you nothing, and it keeps the joke honest, since the scene is a remake and not a recording of you. Platforms keep changing their own labelling tools and rules, so check the current settings of the one you post to.
videosi is not affiliated with Obsession, Curry Barker, Focus Features or Universal Pictures, so do not present the clip as official.
Can I make this with ChatGPT, Sora or Veo instead?
You can try, but it is a different job. A general text-to-video tool starts from words, so you would have to describe five separate shots, keep your face consistent between the booth close-up and the standing shot, and add the dialogue yourself. The timing of the laugh and the tears is also left to chance.
The nice date AI page starts from the original scene as a motion reference, so the shots, the cuts and the 15-second length follow the film, and your photo replaces the woman. You press Generate; there is no prompt to write.
Where does the line come from, and what do reports say about the film?
Obsession is a horror film by Curry Barker. Fangoria reported an official release date of 15 May 2026, with Focus Features distributing in the US after its Toronto festival premiere and Universal Pictures International handling most overseas territories. Dread Central reported a digital release on 30 June.
We only describe the one restaurant scene that the trend uses. The rest of the film is not covered here, and the page does not need it.
What does it cost, and what if it goes wrong?
It is a paid feature, with no trial render and no complimentary generation. A standard 15-second video at 720p is 13 credits, and the Single pack is $6.99 for 10 credits. Higher models and resolutions cost more.
There are no refunds on purchases, but a failed, timed-out or blocked render returns its credits automatically. Check your photo before you press Generate, because that is where most disappointing results start.
Ready to try it? Open the nice date AI page, add your photo and press Generate.
