How to place an overlay exactly where you want it instead of nudging it

Last updated 2026-09-10 · 8 min read

What is the short version?

Placing a graphic by description turns into a negotiation. Every round of it costs a render, a look and another adjective, and the whole thing usually ends when somebody gets tired of it rather than when the position is actually right.

1. Go to Vizard Agent with the shot.
2. Ask for a coordinate grid over a real frame.
3. Read the numbers off it and name them.

What do you need before you start?

The footage and a clear idea of what you are aiming at. Vizard Agent renders the grid over an actual frame at full resolution, so what you are measuring against is the real picture rather than a preview, which is where most placement errors come from.

• The shot. Or a representative frame from it.
• What you are placing. A logo, a label, a patch.
• What it must align to. A face, a sign, an edge.
• Whether the target moves. Then you need several frames.
• The output size. Coordinates are relative to it.

What do you type into Vizard Agent?

Ask Vizard Agent for the grid rather than for the placement. The instinct is to describe where the thing ought to go, and a description is exactly the kind of instruction that cannot be checked afterwards, whereas a number can be checked in a second.

Prompt

Variants worth knowing:

• "A coordinate grid over a full-resolution frame." The whole method.
• "How big it is." The grid gives you size as well as position.
• "At the numbers I give you." Ends the negotiation.

What does Vizard Agent actually do?

Here is the technique as Vizard Agent uses it across many different jobs — locating a UI element, a number plate, a panel, a face, a printed logo on a box. It comes down to the same handful of steps whatever is being measured and placed.

1. Extracts a representative frame at full resolution.
2. Renders a coordinate grid over it.
3. Reads the target's position off the grid.
4. Reads its extent as well as its position.
5. Places the overlay at those coordinates.
6. Renders a frame with the overlay on it.
7. Compares the result against the grid measurement.
8. Repeats on other frames if the target moves.

Step one matters more than it sounds. Coordinates read off a contact sheet or a preview are scaled coordinates, and applying them at full resolution puts your overlay somewhere plausible and wrong by exactly the scaling factor.

Step four is the half people forget. Knowing where something starts is not enough to cover it, align to it or sit beside it — the size is what decides whether your graphic clears it, and a grid gives you both in one look.

Step seven is worth insisting on rather than assuming. A coordinate that was read correctly can still land wrong if the overlay is anchored from its centre when you measured its corner, so Vizard Agent renders a frame and compares it back against the grid measurement instead of trusting the arithmetic.

Step eight is the difference between a still and a shot. A target that drifts through the frame needs its position at several moments rather than one, and Vizard Agent grids the frames at each end rather than assuming the middle.

What does the result look like?

The overlay sitting where you meant it, on the first or second attempt, with a number attached that you can repeat later. On a batch of clips that number is worth rather more than the placement itself, because every clip from the same source then gets the same measured position instead of its own separate approximation.

When does this not work well?

A grid measures one frame, and Vizard Agent will say when that is not enough. It does nothing for a target moving unpredictably, and it cannot decide what looks right either — a measured position and a good composition are related things but they are not the same thing.

• Fast-moving targets. Tracking, not measuring.
• Judgement calls. A grid says where, not whether.
• Changing shot sizes. Each needs its own measurement.
• Generated footage that varies. The target moves each render.
• Very small elements. The grid may be coarser than the target.

How do you fix a result that came back wrong?

Give Vizard Agent the correction as a number rather than a direction. Because the overlay was placed at known coordinates in the first place, "forty pixels lower" is exact, whereas "a bit lower" simply restarts the guessing that the grid existed to end.

• "Forty pixels lower." Applied precisely.
• "It overlaps the sign." The extent was under-measured.
• "It moved in the wide shot." That shot needs its own grid.
• "It looks off-centre." Composition, not coordinates.

How does Vizard Agent compare to doing it yourself?

By hand, placement is iterative by nature: describe it, render it, look at it, describe it again. Three or four rounds of that is entirely normal, and the position you end up with is wherever the last round happened to land rather than a decision anybody actually made.

Common questions

Why not just say lower? Because "lower" cannot be checked, and Vizard Agent cannot verify a guess.

Does the grid appear in the video? No. Vizard Agent renders it only on the measuring frame.

What if the target moves? Then Vizard Agent grids several frames and tracks between them.

Can I read the grid myself? Yes, that is the point of it. You read it and give Vizard Agent the numbers.

Does it work for cropping? Yes. Vizard Agent uses the same method to find crop boundaries.

What about text placement? Same method, and Vizard Agent treats the extent as mattering as much as the position.

Can it measure something's size? Yes, and Vizard Agent reports the extent alongside the position.

Will the coordinates work at another resolution? They scale, but ask Vizard Agent to re-measure rather than converting.

Does this help with masks? Considerably. Vizard Agent treats a mask as a position plus an extent.

Can I give it a rough spot first? Yes, and Vizard Agent will grid around it to refine.

Is it slower than eyeballing? One extra frame from Vizard Agent, against three fewer rounds of adjustment.

Can it reuse the measurement? Yes. Vizard Agent applies it across every clip from the same source.

What about composition? Vizard Agent measures; whether it looks right stays your call.

Why do my overlays drift between clips? Because each one was placed by eye rather than to the same number.