How to make space between two people in a photo without redrawing them

Last updated 2026-09-05 · 8 min read

What is the short version?

A good photograph of two people rarely has room for a title between them. Cutting them out and re-staging them on a new background solves that and loses the photograph. Sliding the real halves apart and rebuilding only the seam keeps it.

1. Go to Vizard Agent with the photograph.
2. Say you want space between them for text.
3. Say the picture should stay as original as possible.

What do you need before you start?

The photograph at full resolution and a sense of how much room the text needs. Everything else is a decision you can make from what comes back — the widths are rendered for comparison rather than argued about beforehand.

• The original photo. The largest version you have.
• What goes in the gap. A title, a logo, a date.
• How authentic it must stay. Say "minimal invention".
• Where it will be seen. A thumbnail is judged small.
• Any logo or badge and how it should sit.

What do you type into Vizard Agent?

Say what you want to keep, not only what you want changed. "Keep as much of the original image authentic" is the instruction that rules out the easy answer — cut them out, generate a new scene, place them in it — which produces a picture that is no longer a photograph of that moment.

Prompt

Variants worth knowing:

• "Keep it authentic." Rules out re-staging.
• "Seamless." The join is the whole job.
• "Give me two widths." Decide by looking.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real podcast thumbnail whose two guests were standing far too close together for a title. The earlier attempts in the session cut both of them out of the picture entirely; the version that finally worked did not touch them at all.

1. Measures how closely a clean plate matches the original.
2. Moves the two halves apart using the photograph's own pixels.
3. Inspects the seams where the halves now meet the gap.
4. Slides the halves apart and blends the middle strip.
5. Compares the plain join against the blended one.
6. Renders the layout into the new gap to see it in use.
7. Softens the grade so the rebuilt strip matches its neighbours.
8. Rebuilds the centre strip two different ways to compare.
9. Tries a minimal-invention version that adds as little as possible.
10. Zooms into the retouched areas looking for marks.
11. Checks a suspicious mark at full resolution.
12. Renders both gap widths and compares them side by side.
13. Exports the retouched photographs for you to work from.

Step two is the difference between this and the obvious approach. Moving the halves means the people, their clothes and the light on them are still the original photograph — only a narrow strip of background is new, and background is the easiest thing in any picture to rebuild convincingly.

Step nine is the standard being aimed at. Every extra invented pixel is a chance for something to look wrong, so the version that adds least is built and compared rather than assumed to be worse than the elaborate one.

What does the result look like?

The same photograph with the two people slightly further apart and a clean strip of background between them. The faces, the clothes and the light are untouched, and the seam is invisible at the size the image will actually be seen.

You get the retouched file, so you can lay your own text over it.

When does this not work well?

The gap has to be fillable in the first place. Two people overlapping each other, or a busy patterned background running behind them, gives Vizard Agent a pattern to continue across the new strip that will not survive close inspection at full size.

• Overlapping subjects. There is no clean line to split on.
• Busy backgrounds. Patterns are hard to continue.
• Someone's arm across the join. It has to be rebuilt.
• Low-resolution photos. The retouching shows.
• Very large gaps. More invention, more risk.

How do you fix a result that came back wrong?

Point at the join and say what you can actually see there. Vizard Agent keeps both centre-strip versions it built and both gap widths it rendered, so preferring one of them is a choice you make rather than another full round of retouching.

• "I can see the seam." The blend rebuilt and re-checked.
• "Too much space." The narrower width used.
• "There is a mark here." Inspected at full size and cleaned.
• "It looks generated." The minimal-invention version instead.

How does Vizard Agent compare to doing it yourself?

The usual approach is to cut both people out and put them on a new background, which is quicker and produces a composite rather than a photograph. Sliding the halves apart is more work and keeps everything that made the picture worth using.

Common questions

Will it cut them out? Only if you want. Vizard Agent prefers to keep the photo intact.

How wide can the gap be? As wide as Vizard Agent can rebuild the background convincingly.

Can I choose the width? Yes. Vizard Agent renders two and you pick.

Will the seam show? Vizard Agent checks at full resolution and at thumbnail size.

Can it put my title in the gap? Yes, or hand you the clean file to do it yourself.

What about the logo? Say where it goes. Vizard Agent keeps its shape as supplied.

Can it fix a halo around a cutout? Yes. Vizard Agent diagnoses it against the backdrop behind.

Does it work with three people? Yes, though Vizard Agent finds each extra join adds risk.

Can it move them closer instead? Yes. Vizard Agent uses the same method in both directions.

Will the lighting still match? Yes, because the light on the people is the original.

Can I get the file to edit myself? Yes. Ask Vizard Agent to export the retouched photograph.

Does it work for a video frame? Yes for a still pulled from one. Vizard Agent finds a moving shot much harder.

What if the background is a plain wall? Then this is easy — Vizard Agent rebuilds plain backgrounds perfectly.

Why not just generate a new scene? Because the point of the photograph is that it happened, and a generated backdrop quietly throws that away.