How to Put Two People in One AI Video (From 2 Photos)

To put two people from two different photos into one AI video, you need a tool that accepts two photos and either merges them into one frame first or passes both to the video model as separate references. On VdoBloom, the Kissing, Hugging, Couple Dance and AI Handshake tabs all have a 1-photo / 2-photo switch. In 2-photo mode the standard models merge both photos into one still frame before animating it, while Seedance 2, Veo 3 Fast and Gemini Omni take both photos directly as separate references.

Disclosure: VdoBloom is our platform.

Why two separate photos are harder than one

Most image-to-video models animate exactly one starting frame. Give them a photo of you and they make you move. Give them nothing about a second person and they either invent one or leave that person out. This is why so many "AI kiss" or "AI hug" attempts end up with a stranger's face, or with one person morphing into the other halfway through the clip.

There are two honest ways to solve it:

Merge first, then animate. An image model takes both photos and composes one new still that shows the two people side by side, with matched lighting and perspective. A video model then animates that still. The video model only ever sees one frame, but both faces are in it.
Multi-reference video models. Some newer video models accept several reference images at once and can be told "the first person comes from image 1, the second from image 2". Seedance 2, Veo 3 and Gemini Omni work this way.

A third route, one photo that already shows both people together, is the easiest of all. If you have a photo of the two of you, use it in 1-photo mode. It skips the merge step, so you get the fewest identity errors.

Which VdoBloom tab to use for each scene

Scene	Tab	What 2-photo mode does	1-photo mode
Kiss	Kissing Video	Merges both photos into one frame, then animates the kiss	One photo that already contains two people
Hug	Hugging Video	Merges both photos into one frame, then animates the hug	Your photo plus an AI-generated partner
Couple dance	Couple Dance	Merges both photos into one frame on the standard models; Seedance 2, Veo 3 Fast and Gemini Omni take both photos as references	One photo containing two people
Handshake	AI Handshake	Merges both photos into one frame on the standard models; Seedance 2, Veo 3 Fast and Gemini Omni take both photos as references	The person in your photo shaking hands

Each tab has preset styles. Kissing offers Romantic, Passionate, Sweet and Dramatic. Hugging offers Warm, Emotional, Gentle and Joyful. Couple Dance offers Waltz, Salsa and Slow Dance. Handshake offers Business Handshake and Deal Sealed. Picking a preset fills in the prompt, and you can edit it before generating. Related tabs such as Lift & Carry, Bridal Carry, Carry Me and Fight use the same 1-or-2-photo switch.

Step by step: kissing or hugging video from two photos

Open the Kissing or Hugging tab and set the photo count to 2.
Upload Person 1 Photo and Person 2 Photo. You will be asked to confirm that everyone pictured is an adult and that you have the right and consent to use the images.
Pick a preset style, or edit the prompt.
Choose a model. These two tabs use a smaller model list (Runway, Kling 2.5 Turbo, PixVerse and Vidu), because the merge step only feeds into those.
Choose 5 or 10 seconds and generate. Behind the scenes, the two photos become one composed frame of both people, and that frame is animated.

What it costs (the merge step is not billed separately):

Runway Gen-3, 5 seconds, 720p: 10 credits. That is exactly the 10 free credits a new account gets.
Runway Gen-3, 5 seconds, 1080p: 20 credits. 10 seconds at 720p: 20 credits.
PixVerse v5, 5 seconds, 720p: 25 credits. Kling 2.5 Turbo, 5 seconds: 40 credits.

For prompt ideas and style-by-style examples, see our full guides on how to make an AI kissing video and how to make an AI hugging video.

Step by step: couple dance, handshake and the other two-person tabs

Couple Dance, AI Handshake, Lift & Carry, Bridal Carry, Carry Me and AI Fight have two routes in 2-photo mode. On the standard models (Runway, Kling 2.5 Turbo, PixVerse, Vidu and LTX), VdoBloom first merges the two photos into one frame, with the first photo’s person on the left and the second photo’s person on the right, and then animates that frame with your prompt. The multi-reference models skip the merge and receive both photos directly. The merge step is not billed separately. Costs for a 5-second clip:

Runway Gen-3, 5 seconds, 720p: 10 credits, which fits in the 10 free signup credits.
Seedance 2 Mini, 5 seconds, 720p: 17 credits (8 credits at 480p).
Seedance 2 Fast, 5 seconds, 720p: 50 credits (24 at 480p).
Seedance 2, 5 seconds, 720p: 82 credits.
Veo 3 Fast: 25 credits per clip, always landscape (16:9) with two photos. Veo 3 Quality is not offered in 2-photo mode. Gemini Omni: 60 credits for an 8-second clip.

On the merged route, describe the action by position, because the video model sees one frame rather than two named people. For example: “the person on the left lifts the person on the right into a bridal carry”.

With Seedance, VdoBloom automatically adds "Reference @Image1 for the first person… reference @Image2 for the second person" to the start of your prompt, so the model knows which face belongs to which person. You can write your own @Image1 and @Image2 tags if you want control, for example "@Image1 leads, @Image2 spins under his arm".

If you only have one photo that shows both people, use 1-photo mode instead and every model in the list works on it.

Photo tips that decide whether the faces survive

Most failures come from the input photos, not the model. Before you spend credits, check the following:

Clear, front-facing faces. No sunglasses, hands over the face, heavy filters or extreme side profiles. The model can only keep what it can see.
Similar framing. Two waist-up photos merge better than a close-up selfie paired with a full-body shot from across the room.
Similar light. A daylight photo next to a dark nightclub photo gives mismatched skin tones in the merged frame.
One person per photo in 2-photo mode. Crop out friends and bystanders, or the model may pull them into the scene.
Reasonable resolution. Very small images get enlarged automatically, but a sharp source always holds identity better. If yours is soft, sharpen it first. Our guide to fixing a blurry image covers how.
Keep the prompt about motion. The presets already tell the model to keep faces, clothing and identity. Add movement and mood ("slow lean in, soft window light"), not new descriptions of the people.

If a face still drifts, try again with the same photos. Each generation is a new attempt. Moving to a higher-tier model (Kling 2.5 Turbo on the kiss and hug tabs, or Seedance 2 on the others) usually holds identity better than the entry models.

Putting two real people into a video they never filmed is exactly the kind of content that can hurt someone. Keep to these rules:

Only use your own photos, or photos of people who have agreed to this specific use. A partner who says yes to a hug video has not agreed to a kissing video.
Never use photos of minors. VdoBloom requires you to confirm that every uploaded person is 18 or older, and it blocks content involving minors.
Never make intimate or sexual content of a real person. VdoBloom blocks non-consensual content, and in many countries making or sharing it is a crime.
Do not use celebrities, politicians or other public figures in videos that could fool people into thinking an event really happened. A fake "handshake" between two real public figures is misinformation, whatever the intent.
If you post the result, label it as AI-generated. YouTube and TikTok both expect realistic AI content to be disclosed.

Frequently asked questions

Is there an AI kissing video generator that works free with 2 photos?

New VdoBloom accounts get 10 free credits. That covers one 5-second 720p kissing video from two photos on Runway Gen-3, which costs exactly 10 credits. After that, clips cost credits. Tools that advertise unlimited free kissing videos usually add watermarks, cap the resolution, or have no clear rules about whose photos you upload.

How do I make an AI video of two people hugging?

Open the Hugging tab, set the photo count to 2, upload one photo of each person, pick a style (Warm, Emotional, Gentle or Joyful) and generate. With only one photo, the tab creates an AI partner to hug you instead.

Can I make an AI hug video without logging in?

Not on VdoBloom. You need an account, because generations cost credits and uploads of real people require an adult-and-consent confirmation. Signing up gives you 10 free credits.

What is a good AI handshake video prompt?

Keep it about the action and setting: "two people giving a confident professional handshake in a bright office, both smiling, slow push-in, cinematic". The Business Handshake and Deal Sealed presets are good starting points. With two photos on Seedance, add "@Image1 on the left, @Image2 on the right" if the positions matter.

Why does the second person's face change or disappear?

Usually the second photo was unclear or at a very different angle and lighting from the first, so the merge could not keep that face. Use sharp, front-facing photos with only one person in each and similar light. If a face still drifts, try a Seedance 2 model, which keeps each photo as its own reference.

Can I make a couple dance video from one photo?

Yes. If one photo already shows both of you, set the photo count to 1 on the Couple Dance tab. This is often the most reliable option, because there is no merge step for the model to get wrong.