How to Make an AI Kissing Video from Two Separate Photos

To make an AI kissing video from two separate photos, open the Kissing tab on VdoBloom, switch to two photos, upload one clear portrait of each person, pick your aspect ratio and a style, and generate. VdoBloom first builds one still frame with both people in the same scene, in the shape you chose, then animates the kiss. The combined frame is free. You pay only for the video, from 8 credits.

Most couples, long-distance partners and fan-fiction writers do not have one good photo of the two people together. That is why they want a kissing video. The hard part is not the kiss. It is getting two people from two different photos, with different lighting, distances and backgrounds, into one believable scene before anything moves. This guide shows what VdoBloom does for you and what you can do to help.

How two-photo kissing works

Older tools just sent both photos straight to a video model. The model tried to merge and animate at the same time, and the usual results were a split-screen look, one person twice, or a stranger's face by the end. VdoBloom splits the job into two steps:

1. Compose. WAN 2.7 places Person 1 and Person 2 together in one scene, keeping each face, hair colour and outfit from your uploads. The frame is made at the aspect ratio you picked (9:16, 16:9 or another supported shape), so the video model never has to crop or stretch it.
2. Animate. The video model you chose turns that one frame into the kiss, with a built-in prompt that tells it to keep both identities and add no extra people.

Because the first step makes an ordinary image, every image-to-video model in the picker works with two photos: Runway, Kling, PixVerse, Vidu, Seedance and the rest.

Step by step

1. Open the tab. Go to Kissing Video and switch the photo count to two.
2. Upload Person 1 and Person 2. One person per photo, face clearly visible.
3. Choose the aspect ratio first. 9:16 for Reels, TikTok and Shorts; 16:9 for YouTube, slideshows or a TV. The combined frame follows this choice.
4. Pick a style. Romantic, Passionate, Sweet or Dramatic. Each loads an editable prompt. You can add a setting, such as "on a rooftop at night", or a camera move, such as "slow push-in".
5. Pick a model and length. Seedance 2 Mini at 480p (8 credits for 5 s) or Runway Gen-3 at 720p (10 credits for 5 s) are good first tries.
6. Generate. The finished clip appears in My Creations, usually within a few minutes.

Choosing photos that combine well

The compose step is good at fixing differences in lighting and background. It works best when the two photos are reasonably similar in these ways:

Check	Why it matters	Quick fix
Similar framing (both head-and-shoulders, or both waist-up)	Very different framing makes one person look too big or too small	Crop the wider photo before you upload
Face turned toward the camera or slightly to the side	Faces in full profile are hard to keep the same once they move	Choose a three-quarter view if you have one
Sharp, well-lit faces	Blurry or dark faces are the most common cause of drift	Avoid night shots, sunglasses and heavy filters
Simple background	Busy backgrounds can leak into the shared scene	A plain wall or outdoor shot works best
Reasonable resolution	Tiny crops lose detail when enlarged	Use the original photo, not a screenshot

If one photo has a transparent background (a cut-out PNG), that's fine: the compose step paints a full scene behind the person, so nobody ends up floating on black.

Height, distance and who leans in

A kiss looks wrong when the heights don't fit. The compose step makes a natural guess from the photos. You can steer it by adding a short line to the prompt, for example "Person 2 is taller and leans down" or "they stand close, foreheads nearly touching". Keep it to one sentence. Long prompts compete with the built-in instructions that keep identities stable.

Which model to finish on

Draft on a cheap model until the combined frame and style look right, then run the same setup once on a model with better motion:

• Seedance 2 Mini, 480p (8 credits): the cheapest way to check that both faces look right.
• Runway Gen-3, 720p (10 credits): a dependable everyday clip.
• PixVerse v5 (25 credits): livelier expressions.
• Kling V2.5 Turbo (40 credits): smoother, more natural movement.
• Runway Gen-4.5 (48 credits): the final render when it really matters.

When it goes wrong

• One person looks like the other. The two photos are too alike in pose and lighting, or one face is small. Use a closer crop of the weaker photo.
• The frame looks squashed. Check that you chose the aspect ratio before generating. The frame follows it.
• A third person appears. Remove crowd scenes from the upload. Use photos with one person each.
• The generation failed. Your credits are returned automatically. Try again with the same photos, or switch model.

For more fixes, see why your AI kissing video looks wrong.

Consent comes first

Two-photo mode makes it easy to put two people together who were never photographed side by side. Only use photos of adults who have both agreed to appear in the video. Never use photos of an ex, a celebrity, a coworker or anyone who hasn't said yes. Your uploads and results stay in your private library.

Common questions

Does combining the photos cost extra?

No. The combined frame is included. You pay only the video model's price, which is shown before you generate.

Can I use a drawing or a game character?

Yes. Original characters and your own artwork work well, as long as they are clearly adults.

Do the clothes stay the same?

The built-in prompt tells the model to keep each person's outfit from their photo. You can change it in the prompt if you want matching outfits.

Can I make the video longer than 5 seconds?

Yes. Most models offer 10 seconds, and the price doubles roughly with length. For example, Runway Gen-3 at 720p is 10 credits for 5 s and 20 credits for 10 s.

Try it now in the AI Kissing Video generator. New accounts get 10 free credits.