How to Make an AI Carry Me Video from Two Photos

To make an AI carry me video on VdoBloom, open the Carry Me tab, upload either one photo containing two people or two separate person photos, choose the Piggyback Ride or Fireman Carry template, pick a model and duration, and generate — a 5-second 720p clip on Runway Gen-3 costs 10 credits ($0.50), and every new account starts with 10 free credits. The effect animates one person picking the other up and carrying them, keeping both faces from the photos you supply.

What the Carry Me effect actually does

Carry Me is a two-person photo-motion effect. It is not a pose filter and it is not a face swap. The model reads your still image or images, works out where the two bodies are, and generates a short video in which one person lifts and carries the other — on their back for a piggyback, or over the shoulder for a fireman carry.

The AI invents the entire carry: the grip, the weight shift, the clothing folds, the background parallax. It works from one frame, so it has no idea what the back of a head looks like. That is why the templates are tuned for short, readable motion rather than a ten-second stunt. Hands and fingers are the usual failure point, and a carry involves a lot of hands.

You must have the consent of every real person whose photo you upload before you animate them into a carry video.

The photo that works best

The Carry Me tab supports one photo or two, and the toggle changes which prompt the tab sends. With a single photo containing both people, the prompt used is along the lines of: someone carrying the person in the photo on their back in a fun piggyback ride, playful and joyful, cinematic quality. With two separate photos, VdoBloom sends the matching two-photo variant instead, so each uploaded person is treated as a distinct subject.

• Full body or three-quarter body, not a headshot. The model needs legs and arms to animate a lift. A chest-up crop gives it nothing to work with and usually produces a floating torso.
• Both faces visible and unobstructed. Sunglasses are survivable; a hand across the face is not.
• Clean separation between the two people. If arms are already tangled in the source photo, the model will invent a fourth arm.
• Even lighting. Heavy backlight or one face in shadow causes identity drift as the body rotates.
• Simple background. Busy patterns behind the subjects warp as the camera moves.

If you are uploading two photos, match them roughly on lighting and framing. Two shots taken in wildly different colour temperatures — one warm indoor, one cold daylight — will composite into a clip that looks like two separate videos stitched together.

The two templates

The Carry Me tab ships with two templates, each carrying its own tuned prompt for single-photo and two-photo mode:

• Piggyback Ride — a fun piggyback carry between two people. The tuned prompt reads: someone carrying the person in the photo on their back in a fun piggyback ride, playful and joyful, cinematic quality. This is the safer of the two. Both faces stay near the camera and the motion is small.
• Fireman Carry — a strong over-the-shoulder carry. Its prompt is: someone carrying the person in the photo over their shoulder in a fireman carry, strong and dynamic, cinematic quality. More dramatic, and more likely to distort the carried person’s face because it rotates them out of the camera plane.

You can also ignore the templates and write your own prompt — the tab’s default is: one person carrying another person in a fun and playful way, cinematic quality. Keep custom prompts short and physical. Describe the carry, the mood and the light. Do not describe faces; the photo already handles identity, and face adjectives in the prompt fight the reference.

Exact credit cost per model

Effect videos are billed at the underlying model’s normal rate — there is no effect surcharge. Credits convert at $0.05 each on the Lite plan ($15 for 300 credits). These are the real numbers from the pricing table:

Model	Tier	Credits	Cost	Best for
Runway Gen-3	720p / 5s	10	$0.50	First attempt; the cheapest honest test
Runway Gen-3	1080p / 5s	20	$1.00	The keeper once composition is right
Runway Gen-3	1080p / 10s	30	$1.50	Longer carry with a walk-off
PixVerse v3.5	720p / 5s	18	$0.90	Cheap second opinion on motion
PixVerse v4.5	1080p / 5s	30	$1.50	Style and camera-movement control
PixVerse v6	720p / 5s	45	$2.25	Best body coherence; billed 9 credits per second
Vidu Q3 Turbo	1080p / 5s	30	$1.50	Audio on the clip
Kling V2.5 Turbo	5s	40	$2.00	Hardest physics; best grip realism
Runway Gen-4.5	5s	48	$2.40	Final hero shot

PixVerse v6 is priced per second rather than per tier — 9 credits per second at 720p, 18 at 1080p — so a 5-second 720p clip is 45 credits and an 8-second one is 72. Runway Gen-3 at 720p for 5 seconds is the only carry configuration that fits inside the 10 free signup credits, which is exactly why it is the tab’s default.

The exact steps

1. Go to /dashboard/video-creation/carry-me/.
2. Choose one photo or two using the photo-count toggle. One photo must contain both people; two photos means one person each.
3. Upload. The tab labels the slots Upload Person 1 and Upload Person 2.
4. Pick Piggyback Ride or Fireman Carry. The prompt box fills with that template’s tuned text — leave it alone on your first run.
5. Pick a model. Start on Runway Gen-3 at 720p / 5s for 10 credits.
6. Set aspect ratio. Use 9:16 for Reels, TikTok and status posts; 16:9 only if it is going on a landscape timeline.
7. Generate, then judge one thing: did the lift read as a single continuous motion? If yes, re-run the same setup at 1080p. If no, change the photo, not the prompt.

In two-photo mode the model roster narrows. Only models that can genuinely consume two reference images stay on the list — the native effect models plus the Seedance 2 family, Gemini Omni and Veo 3. Models whose second image means end frame are deliberately hidden, because they would morph person one into person two instead of carrying them.

A worked example

Two friends, one full-body photo taken outdoors, 9:16. First run: Piggyback Ride, Runway Gen-3, 720p, 5 seconds — 10 credits. The lift read correctly but the background shimmered. Second run: identical settings at 1080p / 5s — 20 credits, background stable, finished. Total spend: 30 credits, or $1.50. That is the normal shape of a good session: one cheap test, one paid keeper.

When to pick a different effect

Carry Me is the fun, casual carry. Other tabs handle neighbouring ideas better:

• Bridal Carry for the romantic front-facing lift, including a Sunset Carry template at golden hour.
• Lift and Carry for Playful, Dramatic or Sweet variations on a straight lift.
• Couple Dance when you want sustained two-person motion rather than a single action.

Paid plans download watermark-free with commercial rights; see pricing for the plan tiers, or the no-watermark generator page for how that works. If you just want to try the effect first, the free AI video generator page covers what the 10 signup credits get you. Model specs live on the Runway Gen-3 page and the PixVerse v6 page.

Frequently asked questions

Can I make a carry me video from one photo?

Yes, if that photo contains both people. The Carry Me tab has a photo-count toggle and sends a different tuned prompt in single-photo mode. If your photo has only one person, the model has to invent the second person entirely, and results are much less reliable — use two photos instead.

How much does one carry me video cost?

10 credits ($0.50) for 720p / 5s on Runway Gen-3, rising to 48 credits ($2.40) for Runway Gen-4.5 at 5 seconds. New accounts get 10 free credits with no card, which covers exactly one 720p test clip.

Why do the hands look wrong?

A carry puts four hands in close contact, which is the hardest thing for any current video model. Reduce it by choosing Piggyback Ride over Fireman Carry, using a source photo where both pairs of hands are clearly visible and separated, and keeping the duration at 5 seconds.

Does the carried person keep their face?

Mostly, and best in piggyback mode where the face stays toward camera. Fireman Carry rotates the carried person away from the lens, so identity drift is more common there. Shorter durations preserve likeness better than longer ones.

Can I use it commercially?

Paid plans include commercial rights and watermark-free downloads. You still need the consent of every real person appearing in the video, and that is on you, not the platform.