How to remove the tags still hanging off the clothes in your video

Last updated 2026-09-17 · 8 min read

What is the short version?

You filmed the shoot before the tags came off. They are small, they dangle on a length of string, and they move independently of the person wearing the clothes — which is exactly what makes them awkward to remove, and why Vizard Agent treats each one as a moving object rather than a spot on the frame.

1. Tell Vizard Agent which items the tags are on.
2. Say the cap, the shoes and the clothing itself are untouched.
3. Expect it to track each tag rather than patch a region.

What do you need before you start?

The clip and a list of the items. Vizard Agent will find the tags itself, but naming the cap and the shoes separately saves a pass, and mentioning where a tag first becomes visible is useful when it only appears once the person turns.

• The clip. The original file.
• Which items. Cap, trainers, jacket, sleeve.
• What the tag looks like. Colour, card, sticker, string.
• What must stay. The garment, the logo, the stitching.
• Where it first shows. Often mid-clip, on a turn.

What do you type into Vizard Agent?

Name the items rather than the frame. "The bottom left corner" stops being true the moment the person moves, whereas "the tag hanging off the trainer" stays true for the whole clip and gives Vizard Agent something to track.

Prompt

Variants worth knowing:

• "Hanging off the cap and the shoes." Two objects, both tracked.
• "Only visible when he turns." Saves a whole missed pass.
• "The floor stays as it is." What the fill has to match.

What does Vizard Agent actually do?

Here is the Vizard Agent sequence on a clip with a card tag on a cap and a beige tag on a trainer. A general removal pass runs first, leaves residue, and the rest of the job is the careful version.

1. Extracts an overview of the frames to find the objects.
2. Enlarges the cap and the shoes at several moments to see what is actually hanging there.
3. Checks which removal tools are available before choosing a method.
4. Runs a targeted removal of both tags, then restores the original duration and audio.
5. Compares original and result at the moments the cap and shoes appear.
6. Runs a motion check on the uploaded copy and enlarges the timestamps it flags.
7. Finds the shoe tag still visible and switches to a local, coordinate-based removal.
8. Traces each tag's trajectory so the local fill follows its movement.
9. Checks when the shoe tag first appears and when it disappears on a turn.
10. Tests the local fill on reference frames before committing to the clip.
11. Re-runs the test with absolute timings when the first test used the wrong offsets.
12. Samples floor from immediately beside the tag rather than using a generated fill.
13. Shrinks the patch to the tag's measured outline so nothing reads as a rectangle.

Step twelve is the one that makes it invisible. A fill invented for the gap has to guess the floor's tone, its grain and its lighting; material lifted from two centimetres away already has all three, so the repair continues the surface instead of interrupting it.

Step thirteen is the finishing move. Even correct material shows as a shape if the patch is bigger than the thing it covers, so Vizard Agent measures the tag's outline and reduces the patch to it.

Step eleven is a genuine mistake, caught by testing. The first local pass used relative rather than absolute timings, so the masks were applied in the wrong seconds; Vizard Agent found it on the reference frames rather than in the delivered file.

What does the result look like?

The same clip with no tags on anything. The cap keeps its shape and its stitching, the trainers keep their logo and their texture, the floor under the missing tag continues as floor, and the audio and duration are the originals.

When does this not work well?

Some tags cannot be taken off cleanly. A tag held against a patterned fabric, one that overlaps the item's own logo, or one that is in front of a face for part of the clip leaves a repair with nothing convincing to rebuild from.

• Tags over patterned fabric. The pattern has to be invented.
• Tags across a logo. Removing one damages the other.
• Very fast movement. The tag smears across several positions per frame.
• Tags in shallow focus. Their blur does not match the surround.
• Tags that fill the frame in a close-up. Too little context to fill from.

How do you fix a result that came back wrong?

Give the moment and the item in the same sentence. Vizard Agent keeps each tag as its own tracked object with its own fill, so a complaint about the shoe at eight seconds does not trigger a rebuild of the work you have already approved on the cap.

• "The shoe tag is still there at 0:08." That tag's path is re-measured from that point.
• "There is a pale rectangle." The patch is shrunk to the measured outline.
• "The floor looks smudged." The fill is resampled from immediately beside it.
• "The trainer lost its logo." The mask is tightened to the tag alone.

How does Vizard Agent compare to doing it yourself?

By hand the honest options are a reshoot or a blur. A reshoot means getting everyone back, and a blur over a swinging tag means a soft smudge travelling around the frame that draws more attention than the tag did.

Common questions

Can it remove several tags at once? Yes. Vizard Agent tracks each one separately and fills each from its own surroundings.

What if a tag only appears halfway through? Say so, and Vizard Agent checks the whole clip rather than the opening.

Will the shoe still look like the shoe? Yes. The mask is kept to the tag, and the logo and stitching are protected.

Does the clip get shorter? No. Vizard Agent restores the original duration and audio after the removal.

What about a sticker on a garment? The same approach, and usually easier because a sticker does not swing.

Can it do a photo instead? Yes, and a still is far simpler — there is one position to fix.

Why did the first pass leave residue? A general pass treats the whole frame. The local pass is aimed at measured coordinates.

Will I see a patch edge? Not if the fill comes from beside it. Tell Vizard Agent if you can see one.

Can it handle a patterned floor? Sampling from alongside usually continues the pattern, but a strong repeat is harder.

What if the tag is over the person's face? Then it is a different job, and Vizard Agent will say so rather than smudge a cheek.

Does it work on a moving camera? Yes. Vizard Agent measures the tag's path in frame coordinates through the move.

How long does it take? Most of the time goes on tracking, not rendering.

Can it keep a tag on one item and not another? Yes. Name which items, and Vizard Agent leaves the rest alone.

What if I have the same problem on twenty clips? Once the approach is settled, Vizard Agent can repeat it across the set.