How to cut the dead space out of a clothing showcase
Upload the take and tell Vizard Agent to cut everything that is not the product. Vizard Agent transcribes the speech, splits the take into shots, zooms into the garment details to see what is actually being shown, and rebuilds the video with the dead space removed and the captions placed clear of the frame's existing graphics.
What do you type into Vizard Agent?
Say what to remove rather than what to keep. Vizard Agent reads "cut the extra parts" as an instruction to find every stretch where nothing is being shown, which is more reliable than listing the moments you want. Name the items too: the on-screen text names them, and a transcript of casual speech won't always get them right.
Prompt
What does Vizard Agent actually do?
A phone take of clothing is mostly reaching, adjusting and finding the angle, so cutting the dead space is the whole edit. The step people skip is looking closely at the product before choosing which seconds to keep, because the seconds worth keeping are the ones where it is actually visible. In the session behind this article it:
- Probed the video, extracted the speech transcript, and analysed the frames and the speech together, including where the useful speech sat in the audio.
- Split the take into shots, built an overview of every cut, and reviewed all the frames in one view.
- Searched for trendy music and analysed each candidate's structure and rhythm.
- Zoomed into the shirt's print and the trousers' fabric to see what was being shown, then checked the first and last frames of every shot for continuity.
- Sourced a typeface for the caption language, tested it on frame, joined the usable sections, and moved the captions after finding them colliding with stickers already in the picture. The delivered file was 1080x1920, 30fps H.264, 12.97 seconds long.
When does this not work well?
This edit can only keep what the take contains. Vizard Agent finds every usable second and cannot create the shot you didn't get. Colour from a phone is hard to trust, there's no fabric beat if you never held on the fabric, existing stickers constrain where captions go, and a short take stays short. Right-to-left languages also need font work.
Common questions
How short will it end up? As short as the usable footage. Vizard Agent delivered thirteen seconds from a longer take in this run.
What about stickers I already added? Vizard Agent checks the frames and places the captions clear of them.
Can Vizard Agent handle right-to-left captions? Yes. It sources a suitable typeface and tests the rendering on a frame first.