# AI kissing video: kiss, hug and dance clips from two photos — our test and prompts

> An AI kissing video can be made from two separate portraits or one photo of a couple: a video model brings the people into one scene and films a kiss, a hug or a dance with sound. We tested all three on the same fictional couple in AI2Face with Gemini Omni. Below are the clips, full prompts, six ready-made scenes and which photos work best.

Published 2026-10-31 · Updated 2026-10-31
URL: https://ai2face.com/blog/ai-kissing-video

A kissing video from a photo is one of the most popular short-video trends: couples make them for anniversaries, friends for a laugh, creators for AI characters. Ready-made apps give you a one-tap kiss template, while a video model with references lets you film your own scene — the place, the light, the camera move and the sound. We filmed a kiss, a hug and a dance with one fictional couple. Every frame and clip was made in the AI2Face generator, and the people in them are fictional.

## How AI makes a kissing video

There are two kinds of input and two ways to control the scene.

| | Template effect in an app | Your own prompt in a video model |
|---|---|---|
| What you upload | one couple photo or two photos | one couple photo or two separate portraits |
| Scene | the same as in the photo | anything: a bridge, a station, a kitchen, rain |
| Motion | one per template | you describe it: leaning in, a forehead kiss, a spin |
| Sound | template music or none | generated for the scene: a river, a station, a song |
| What changes the result | only the template you pick | every word of the prompt |

**One couple photo** becomes the first frame: the clip starts exactly from your shot. **Two separate portraits** are references: the model places the people in a new scene itself. The second route is for when you don't have a photo together — a long-distance greeting, say, or a couple of AI characters.

## How to make a kissing video from two photos

We took two separate portraits: a man in a denim shirt in a park and a woman in a cream sweater against autumn trees.

![Two portraits side by side: a man with a short beard and dark wavy hair in a denim shirt, and a woman with long dark hair and silver hoop earrings in a cream sweater](https://ai2face.com/_astro/two-photos.CgektKC2_Z1tLX27.jpg "The inputs: two separate portraits of fictional people. Nano Banana Pro")

Both photos went in as references to [Gemini Omni](/models/gemini-omni). The prompt says who comes from which photo, what they wear, where they stand and how the camera moves:

```
Romantic vertical video at golden hour: the man from the first reference photo
(short dark wavy hair, short beard, navy denim shirt) and the woman from the
second reference photo (long dark-brown wavy hair, silver hoop earrings, cream
knit sweater) stand facing each other on an old stone bridge over a river at
sunset. She smiles, he gently tucks a strand of hair behind her ear, they lean in
and share a soft, tender kiss, then rest their foreheads together and smile.
Slow push-in, medium close-up, warm backlight, light breeze. Keep both faces
exactly as in the photos. Sound: river, distant city, soft wind. No text.
```

[A couple on a stone bridge at sunset: he tucks a strand of her hair behind her ear, they kiss and rest their foreheads together](https://ai2face.com/_astro/media/blog/ai-kissing-video/kiss.mp4 "A kiss from two separate portraits: Gemini Omni 1.1 Flash, reference mode, 9:16, 6 s, with sound")

The model played every beat of the scene in order: the smile, the strand of hair, the kiss, foreheads together. The faces and clothes match the portraits, and her earrings are still there. The model set the sunset light on its own — the source photos were taken on an overcast day.

## AI hug video from photos

A hug works the same way — only the action changes. We put the couple on a train platform: she spots him, runs over, and he lifts her and spins her around. The sound is station ambience only: we ruled out laughter and voices in a separate line of the prompt.

```
Vertical video of a reunion at a train station platform in the morning: the
woman from the second reference photo (long dark-brown wavy hair, cream knit
sweater) sees the man from the first reference photo (short dark wavy hair,
short beard, navy denim shirt, a travel bag on his shoulder), runs to him, and
they hug tightly; he lifts her slightly off the ground and spins half a turn,
both smiling, her eyes closed on his shoulder. Handheld camera, medium shot,
soft morning light, a blurred train behind them. Keep both faces exactly as in
the photos. Sound: only station ambience — footsteps, a distant train, soft
murmur of the platform. No laughter, no voices, no speech. No text, no readable
signs.
```

[A woman in a cream sweater runs along a platform to a man with a bag; they hug tightly and laugh in front of a train](https://ai2face.com/_astro/media/blog/ai-kissing-video/hug.mp4 "A reunion hug from the same two portraits: Gemini Omni 1.1 Flash, 9:16, 6 s, station ambience only")

The hug came out livelier than the kiss: there's running, smiling and a half spin. The model added the shoulder bag as asked and kept the platform sign blurred, with no readable text. Describe the sound explicitly, including what it shouldn't have: “No laughter, no voices.”

## AI dance video from a couple photo

If you already have a photo together, set it as the first frame and the clip will start exactly from your shot. We assembled the couple photo from the same portraits in [Nano Banana Pro](/models/nano-banana-pro) — the same way a couple photo is built from two shots in our guide on [how to combine photos with AI](/blog/obedinit-foto-nejroset).

![A couple in a cozy kitchen with string lights: a man in a denim shirt holds a woman in a cream sweater by the waist, both smiling at the camera](https://ai2face.com/_astro/couple-photo.B5TImrLW_27ziGA.jpg "The couple photo in the kitchen — the first frame for the dance. Built from two portraits in Nano Banana Pro")

```
The couple starts a slow dance in their cozy kitchen: he takes her hand, she
laughs, he gently spins her under his arm and pulls her close, they sway together
cheek to cheek under the warm string lights. Static camera at chest height, smooth
natural motion. Keep both faces, hair and clothes exactly as in the image.
Sound: a soft acoustic love song playing from a speaker, her quiet laugh. No text.
```

[A couple in a kitchen under string lights: he takes her hand, spins her under his arm, pulls her close, and they sway cheek to cheek](https://ai2face.com/_astro/media/blog/ai-kissing-video/dance.mp4 "A dance from one couple photo: first frame is the photo above, Gemini Omni 1.1 Flash, 9:16, 6 s, with a song")

The model handled the slow dance with one spin smoothly — no extra hands and no rubbery moves. Fast dances with lifts and jumps are harder for video models: keep the choreography simple and the clips short.

## AI kissing prompts: 6 ready-made scenes

The prompt formula: **who** (which photo, key features) + **where** (place, time of day, light) + **the action** step by step + **camera** + **sound**. The scenes below slot into that formula. Swap in your own people's features.

**Kiss in the rain.**

```
The couple stands under one umbrella on a night city street, he lowers the
umbrella, they kiss in the rain, raindrops in the warm street light. Slow
motion, medium close-up. Sound: heavy rain, distant traffic.
```

**Forehead kiss.**

```
The man gently holds the woman's face in his hands and kisses her on the
forehead, she closes her eyes and smiles. Soft window light in a quiet bedroom,
close-up. Sound: quiet room tone, a soft breath.
```

**Spin and kiss.**

```
On a sunny beach the man lifts the woman and spins her around, puts her down and
kisses her, both laughing. Low-angle handheld camera. Sound: waves, wind, their
laughter.
```

**Cheek kiss.**

```
In a busy café the woman leans across the table and kisses the man on the cheek,
he blushes and laughs. Static camera, medium shot. Sound: café chatter, cups
clinking.
```

**Wedding kiss.**

```
The bride and groom stand under a flower arch outdoors, guests throw petals as
they kiss. Camera slowly rises from medium to wide shot. Sound: applause and
cheers.
```

**Kiss with a pet.**

```
The woman from the reference photo hugs her golden retriever on a sunny lawn,
the dog licks her cheek and she laughs. Handheld camera at eye level. Sound:
birds, her laughter, a happy bark.
```

Video models follow spoken lines and on-screen text less reliably than actions. If you need a line, keep it short and in English: Google officially supports only English in Omni. For couple photos rather than video, see our [couple photoshoot prompts](/blog/promty-dlya-parnoj-fotosessii).

## Which photos work for an AI kissing video

- **Front-facing or slightly turned portraits** with the eyes visible: the model holds a face better when it sees all of it.
- **Even daylight** without harsh shadows or color filters.
- **No sunglasses, masks or hands near the face.**
- **A close face:** waist-up or tighter. In a wide shot the face is small and the likeness slips.
- **The clothes in the photo are the clothes in the clip** unless you describe others. In our test the sweater and the shirt carried into all three scenes.
- **For a couple photo,** both faces visible and not overlapping.

## Limitations and ethics

- **Length.** One Gemini Omni generation runs up to 10 seconds. A longer story is built from several clips.
- **Complex motion.** Fast dances, lifts and intertwined arms produce artifacts more often than a calm kiss or hug.
- **Consent.** Use your own photo and a photo of a partner who's fine with it. In Russia, Article 152.1 of the Civil Code allows using a person's image only with their consent, and your country's rules may differ. Don't make clips with exes, coworkers or strangers.
- **Public figures.** Google doesn't allow uploading images of recognizable public figures to Gemini Omni — a “me kissing a celebrity” clip will be rejected.
- **Watermarking.** Gemini Omni clips carry an invisible SynthID watermark.

## Bottom line

An AI kissing video works best from two sharp portraits and your own prompt: you control the place, the motion and the sound, and the faces stay recognizable. If you already have a photo together, set it as the first frame and the clip starts from your shot. Template effects in apps are faster, but they give everyone the same scene. For an anniversary gift, a long-distance greeting or a clip of an [AI character](/blog/ai-influencer) couple, use a video model with references. The kiss, the hug and the dance are all made in the AI2Face generator, with the cost shown before each run; to animate a single photo, see the guide on [how to animate a photo](/blog/ozhivit-foto).

## FAQ

### How do I make an AI kissing video from a photo?

Upload two portraits or one couple photo to a video model that accepts references, such as Gemini Omni, and describe the scene: where the couple stands, how they come together and kiss, how the camera moves. In about a minute you get a clip with sound.

### Can I make a kiss from two different photos?

Yes. Attach both portraits as references and say in the prompt who comes from which photo. In our test the faces from two separate shots held up in both the kiss and the hug.

### Can I make an AI kissing video for free?

Yes, for a test. AI2Face gives you welcome credits when you sign up — enough for a short clip. The cost is shown before you start, and the service adds no watermark to the result.

### What photos work best for an AI kissing video?

Sharp front-facing portraits in daylight, without sunglasses, masks or heavy filters. The face should take up a good part of the frame.

### Can I make a hug or a dance the same way?

Yes, only the description of the action changes. We made the reunion hug from the same two portraits and the slow dance from a single couple photo in a kitchen.

## Sources

- https://ai.google.dev/gemini-api/docs/omni
- https://blog.google/innovation-and-ai/technology/developers-tools/build-with-gemini-omni-1-1-flash/
- https://blog.google/innovation-and-ai/products/nano-banana-pro/
- https://www.consultant.ru/document/cons_doc_LAW_5142/14c6c3902cffa17ab26d330b2fd4fae28e5cd059/