AI Walk Video with Two People: What Actually Works

The most common two-person request is a single photo of two friends, and the most common disappointment follows it within minutes: the clip comes back with one person walking, or with two faces that melt into each other.

The reason is structural. A render takes one photo and performs one walk with it. Two people in one frame is not a feature that is missing; it is a different problem, and the honest workflow is two renders.

Why one photo cannot do two people

The model needs to decide who is walking, where the body ends, and how the limbs connect. With two faces it has to guess all three, and it guesses differently in frame ten than in frame forty. The result is the melting you have seen.

Give it one person per photo and there is nothing to guess. That is why the photo checklist insists on one subject, and why a group photo is the single most reliable way to waste a credit.

The workflow that works

1. Two photos, one per person. Same lighting conditions if you can, and ideally the same kind of framing: knees-up or full body, one person, facing the camera.

2. Two renders. Each render is one clip. If you want a pair of clips that look like they belong together, keep the tier identical: two 480p clips, or two 1080p clips. Mixing tiers makes one clip visibly sharper, and viewers notice the difference immediately even if they cannot name it.

3. One edit. Put the two clips side by side or back to back in whatever editor you already use. Two twelve-second clips give you a twenty-four second piece, which is a good length for a duet format.

4. One upload. Post the edited video as one item, not as two separate posts. Two posts split the engagement and look like a mistake.

What it costs

Each render is charged separately, so a pair is exactly twice the credits: 20 credits at 480p or 60 at 1080p. A failed render refunds its own credits, so a half-broken pair still returns the half that failed.

That is worth saying plainly before you start, because people often assume a "duo" feature exists at a discount. It does not; you are buying two renders, and the price is the price of two.

If you want to know which pack makes a pair cheapest, the credit maths runs the numbers per pack.

Making the pair feel like a pair

Four small things do most of the work.

Match the framing. If one person is full body and the other is cropped at the chest, the two clips will not sit together.

Match the light. One clip in open shade and one in hard sun reads as two different days. Shoot or pick both photos at similar times if you can.

Order them deliberately. Whoever walks first sets the tone. If one clip is visibly stronger, lead with it.

Keep the sound continuous. One track across both clips, not two tracks stitched together; the seam in the audio is more noticeable than the seam in the video.

Splitting the cost with a friend

If you are making a pair with someone else, decide up front who pays for which render. The tool charges per render, so the cleanest split is "each of us buys one", which also means each of you owns the clip of yourself.

The awkward version is one person buying both and then asking for half afterward: the second render has already been spent, and if the friend's photo turns out to be unusable, you are negotiating over a credit that no longer exists. Buying separately avoids that conversation entirely.

If you do buy both, buy the cheaper tier for the first attempt on each photo. Two test renders at 480p cost the same as one at twelve hundred, and you will know whether both photos work before you spend the higher tier on either.

What two people cannot do

Two people cannot walk toward each other and meet in one clip. They cannot hold hands. They cannot pass something between them. Each of those needs both subjects in the same generated shot, which this tool does not produce.

If a shared moment is the point, shoot it yourself and edit the AI clips around it. Mixing generated walking shots with a real clip of the two of you is a legitimate format, and it is honest about which part is which.

The short version

One photo, one person, one render. A pair is two of those, edited together, posted once. It costs twice as much and looks like a format rather than a mistake, which is the whole point.

Ready to make your half of the pair? Open the studio — 10 credits at 480p, refunded if the render fails.