How to make a video that matches your Figma design

Last updated 2026-09-01 · 8 min read

What is the short version?

Somebody has already decided what your brand looks like, and it lives in a design file. A video that invents its own captions and colours is a second, competing brand — even when it looks perfectly nice on its own.

1. Go to Vizard Agent and upload the footage.
2. Give it the link to your design file.
3. Say to use only that style, and to invent nothing on top of it.

What do you need before you start?

The footage and a design file it can reach. Vizard Agent samples the palette out of the artwork itself, so a file with the actual components in it is far more useful than a written list of hex values that may already be out of date.

• The footage. As filmed.
• The design file. Reachable, with the real components.
• The rule. Use this style only.
• The typefaces. Named, in case they are not installed.
• The platform. It sets the frame and the safe area.

What do you type into Vizard Agent?

Point at the file and then explicitly close the door behind it. Vizard Agent will otherwise improve on the design wherever it spots an opportunity to, and a caption style that is very nearly yours is considerably worse than one that is obviously not yours at all.

Prompt

Variants worth knowing:

• "Only use styles from…" The instruction that binds it.
• Name the typefaces. They may need installing.
• Say what may be added. B-roll and music usually sit outside the design.

What does Vizard Agent actually do?

Here is the order Vizard Agent worked in on a real video built to a design file. Steps nine and eleven are the ones that matter: it samples the colours twice, out of the artwork, rather than reading a value from a note.

1. Looks at the attached image and checks the footage's details.
2. Checks the specifications of both files.
3. Reads the project history for the client's previous work.
4. Transcribes, surveys the shots, fetches the design reference and grabs frames in one pass.
5. Builds a frame contact sheet and reads the shot log.
6. Reviews the footage frames and reads the full transcript.
7. Samples the design's colours and lists the available typefaces.
8. Lists the available effects.
9. Samples the design palette again to pin the values down.
10. Loads the text-behind-subject treatment and reads the caption and graphics options.
11. Measures the speaker's head position in the frame.
12. Confirms the brand name as she actually says it.
13. Checks the head position against the title placement.
14. Pulls word-level timings so the overlays sync to the speech.
15. Generates the overlays, the cutouts, the music and the B-roll in parallel.

Step eleven is what stops a well-branded video from being a badly framed one. A design file positions a title in an empty rectangle; real footage has a person in that rectangle, and the placement has to be measured against where they are.

What does the result look like?

The video this page is written from came back with captions and graphics drawn from the client's own design file — its palette sampled from the artwork, its typefaces matched, its treatments applied — over footage with the dead air removed and the titles placed clear of the speaker's head.

Nothing in it is a near-miss. The blue is the blue from the file rather than a blue that looks like it, which is the difference an art director notices immediately and nobody else can name.

When does this not work well?

A design file describes a still layout sitting on a page, and video introduces two things that layout never had to account for: the passage of time, and a person who moves around inside the frame. These are where the two do not meet.

• Typefaces you do not own. Licensing is yours to check.
• Layouts with no room. The person is in the title's space.
• Styles with no motion defined. How should it appear?
• Very complex components. Some do not survive at video size.
• A file it cannot reach. Export the frames instead.

How do you fix a result that came back wrong?

Point at the component in the design file that is not being respected. Vizard Agent keeps the sampled palette, the measured head positions and all the word timings, so a corrected colour or a repositioned title re-renders without the underlying edit being touched at all.

• "That is not our blue." Re-sampled from the file.
• "The title covers her." Repositioned from the head measurement.
• "Wrong caption weight." Matched to the component's actual weight.

How does Vizard Agent compare to doing it yourself?

By hand this means an editor reading hex values off a screenshot and choosing the nearest installed font — which is how videos end up almost on-brand. Vizard Agent samples the values out of the file and measures the frame before placing anything.

Common questions

Does it need access to the file? It needs to be able to fetch it. Vizard Agent can work from exported frames instead.

Will it get the colours exactly right? Yes. Vizard Agent samples them from the artwork rather than approximating.

What if the typeface is not installed? It can install it, licence permitting. Vizard Agent tells you when it cannot.

Can it use styles not in the design? Not if you forbid it. Vizard Agent stays inside the file's vocabulary.

What about motion? Design files rarely define it. Vizard Agent asks or keeps the motion minimal.

Will the titles cover the speaker? No. Vizard Agent measures the head position and places around it.

Can it match a brand guidelines PDF instead? Yes. Vizard Agent reads the values out of whatever you supply.

Does it do the whole edit too? Yes. Vizard Agent cuts the dead air and adds B-roll within the same pass.

Can it caption in the design's style? That is the main use. Vizard Agent applies the caption component as specified.

What about my logo? Taken from the file. Vizard Agent uses the real asset rather than a screenshot.

Why sample rather than read the hex? Because the note and the file drift apart. The file is what the designer maintains; a hex value written down six months ago is a guess about it.

Can it produce a second video to match? Yes. Vizard Agent keeps the sampled palette and applies it to the next one.

Does it check the result? Yes. Vizard Agent reviews the title placement and the rendered frames before committing to the final render.

Can it work from a Canva file? Yes. Vizard Agent samples the values from whatever design source it can reach.