How to remove a graphic that sits behind the text you are keeping

Last updated 2026-09-22 · 8 min read

What is the short version?

A logo or symbol sits behind the title on screen. Removing it is easy in the areas where it sits alone and genuinely hard where it runs under the lettering, because anything Vizard Agent paints there paints the letters as well.

1. Tell Vizard Agent which graphic to remove.
2. Say the text must remain pixel-identical.
3. Expect it to test the overlap area on single frames first.

What do you need before you start?

The video and a clear statement of what must not change. Vizard Agent treats "keep the text the exact same" as a hard constraint it can verify by comparison, which is much stronger than asking it to be careful.

• The video. The original file.
• The graphic. Which one, and where.
• What stays identical. The text, the composition, the frame.
• Whether it moves. A static overlay or an animated one.
• The output format. Especially if the frame is square.

What do you type into Vizard Agent?

Name the exception twice. "Remove the umbrella symbol behind the text, keep the text exactly the same, keep everything else the same" sounds repetitive and it is what makes the job checkable — two of those three clauses are tests.

Prompt

Variants worth knowing:

• "Behind the text." Where the difficulty is.
• "Exactly the same." A comparison, not a preference.
• "Keep the square composition." Stops a reframe.

What does Vizard Agent actually do?

Here is the Vizard Agent sequence on a square video with a symbol overlapping its title text. Four different approaches to that overlap get tested on single frames before anything at all is rendered across the whole clip.

1. Inspects the video and samples frames.
2. Extracts detailed frames at several moments.
3. Inspects the overlap between the symbol and the text at full resolution.
4. Checks which object-removal options are available and their controls.
5. Removes only the symbol while preserving the square composition and the text.
6. Runs the square-preserving removal at full resolution.
7. Compares source and edited frames at matching moments.
8. Prepares high-resolution title crops and alignment guides for a pixel-preserving cleanup.
9. Inspects the original title pixels and the symbol's boundary.
10. Tests a localised symbol-only cleanup on a source frame.
11. Tests a source-preserving patch that uses the cleaned background only under the title.
12. Tests texture inpainting from the original frame on the symbol's region.

Steps ten to twelve are the comparison worth understanding. Each approach fails differently: a general removal softens the letters, a patch shows a seam where the background changes, and inpainting invents texture that does not match. Testing them on one frame costs seconds each.

Step eight is the technique that works. Keeping the original title as its own high-resolution crop means the background can be rebuilt freely and the letters put back exactly as they were, rather than being repaired.

Step seven is the verification. "Keep the text identical" is only meaningful if somebody compares the two, and the comparison is done at matching timestamps rather than by eye.

What does the result look like?

The same video with the symbol gone and the title untouched: the letters as sharp as they were, the background behind them continuous, the composition and resolution unchanged, and no soft patch where the graphic used to be.

When does this not work well?

Some overlaps simply cannot be undone. If the symbol is semi-transparent and its colour has already mixed with the letters, or if the type is thin and the symbol busy, there is no clean separation for Vizard Agent to recover.

• Semi-transparent graphics. The colours are mixed into the letters.
• Thin or script type. Its edges share pixels with the symbol.
• Animated symbols. Each frame is a different overlap.
• Heavy compression. The blocking spans both elements.
• Text with a glow or shadow. The effect sits over the symbol too.

How do you fix a result that came back wrong?

Say what changed in the frame that should not have changed. Vizard Agent compares the affected frames against the source file and restores the text from the original pixels rather than trying to repair the damaged letters a second time.

• "The letters look soft." The original title pixels are composited back over the patch.
• "There is a seam." The patch is rebuilt from a wider area of background.
• "The symbol shows in one shot." That range is treated separately.
• "The video was reframed." The square composition is restored.

How does Vizard Agent compare to doing it yourself?

By hand you mask the symbol and inpaint, and the letters that cross the mask come back slightly wrong — slightly rounder, slightly softer. It is the kind of damage nobody can name and everybody notices when the two versions are side by side.

Common questions

Can it remove a watermark behind text the same way? Yes. Vizard Agent treats any overlapping graphic with the same method.

What if the symbol animates? Then each frame is its own overlap, and Vizard Agent tracks it through the clip.

Will the text be re-typed? No. Vizard Agent keeps your original letters rather than recreating them.

Can it keep the exact resolution? Yes. Tell Vizard Agent, because some removal routes resize by default.

What about a square video? Say so, because Vizard Agent will otherwise deliver the platform's usual shape.

Does the audio change? No. Vizard Agent treats this as a picture-only correction.

Can it replace the symbol with another? Yes — the same rebuild, with your graphic composited instead.

How do I know the text is identical? Ask Vizard Agent for the source-versus-edit comparison frames.

What if the background is busy? Then the patch has more to match, and Vizard Agent will test more approaches.

Can it do this on a still? Yes, and it is far quicker with one frame.

Will it work at low resolution? Less reliably, and Vizard Agent will say so — the letters and symbol share pixels.

Can it remove several graphics? Yes, each with its own test and patch.

How long does it take? Most of the time is testing approaches on frames, not rendering.

What if part of the symbol is outside the text? That part is the easy half, and Vizard Agent does it first.

Can it rebuild an animated background underneath? Yes, though a moving background means the patch has to be rebuilt for every frame rather than once.

Should I supply the original title artwork? If you have it, yes — Vizard Agent can composite your own file back over the repaired background instead of lifting the letters out of the video.