Theuws ConsultingCase study ← Back to the video studio guide
Second brand · start to finish

A second brand, made real: Elementals.nl

The same video studio, pointed at a different, live product — Elementals.nl, a Big Five personality assessment. We adopted its existing brand, built a reusable leader, drove its real /en app for the footage, and shipped a finished English walkthrough in both 16:9 and 9:16 — for about $0.61 in API generation. Here is exactly how, and what it cost.

By Theuws Consulting · a making-of · 10 July 2026

"Meet your elements" — 31.5s, 16:9. Generated five-elements leader · real /en screens · a Norse-archetype hero · titles set in the brand's own font. Made by the studio, held to the same quality bar.

≈ $0.61
Total API generation
Adopted
Brand, not invented
16:9 + 9:16
One build, both ratios
Real app
Its own /en screens
01 · Intake

Adopt the brand — don't invent one

Elementals already had a complete style guide, implemented as design tokens. The studio's intake gate adopts that canon rather than guessing: the exact palette, the display font (Outfit), and the tone rules ("assessment", never "test"). The film's titles are set in Outfit too — so they read as part of the product, not stuck on top of it.

Earth · #8B6914 Water · #1B4965 Fire · #C45B28 Wind · #708090 Aether · #7B68AE Base · #1A1614
02 · The signature, proven once

One keyframe, validated before anything scales

The brand mark is a five-sided pentagon — five elements. So the leader is five elemental energies converging into that pentagon. We generated a single keyframe first (nano-banana), checked it, then animated it (image-to-video) and reversed it for the outro. Proving the hardest shot once is what keeps a series cheap — you never fan out a look you haven't confirmed.

Leader keyframe — five elemental wisps converging into the Elementals pentagon on warm near-black
03 · Real footage, reproducibly

We drove the real product — using its own test rig

The screens are not mock-ups. We captured the live /en app deterministically, reusing Elementals' own Playwright setup (the one it runs UATs with) — so every run yields identical frames, and languages or variants can be re-captured without drift. Screen capture like this is free; it is the biggest reason a walkthrough stays cheap.

Three real /en screens captured deterministically: landing, the free example result, and the 16 Norse archetypes

Landing

The real hero: "Discover your Big Five type and archetype in 5 minutes."

The free result

A public example report — Big Five bars + archetype — so no login was needed.

16 archetypes

The Norse-archetype grid, each tagged in its element's colour.

04 · Fan-out

Both ratios from one build — for $0 more

The 9:16 cutdown reuses every paid asset: the leader is reframed (the pentagon sits centre), the screens are re-captured natively on a mobile viewport and scroll through the real pages, and the titles re-flow to portrait. No new generation — so the vertical costs nothing extra to produce.

The vertical is a genuine native-mobile capture — the pages scroll to reveal real content (the hero, the free-report promise, the founder, the archetype cards), not a stretched desktop grab.

The ruleBuild one master, reframe and re-capture for the cutdown. Paid generation happens once; every ratio and language after that is edit work — which is free.
05 · The economics

What it cost — and when it costs more

The honest headline: the money is in AI generation — mainly image-to-video — not in the edit. Screen captures, stills with motion, the montage and every re-run are free. Here is this film, itemised.

ItemHowCost
Leader keyframenano-banana Pro, 2K
Leader motion (→ reversed for the outro)fal Veo-3-Fast, 1080p, 4s
Outroreversed leader (ffmpeg)$0
Archetype herostill + ffmpeg push-in & glow$0
3 real /en screensPlaywright capture rig$0
Audio bed (42s)fal stable-audio
Montage + grade (16:9 and 9:16)ffmpeg + ImageMagick$0
Totalone paid bracket + one bed; the rest free≈ $0.61

It costs more when…

  • more image-to-video shots × more seconds × higher resolution (i2v is by far the biggest line)
  • you use higher-tier models (Pro, Veo-3, 4K) where a cheaper tier would do
  • you fan out to N languages/ratios that each generate instead of reusing one master
  • you iterate without a PoC — regenerating a look you never validated
  • the film is fully-AI: many generated shots instead of captured ones

It costs less when…

  • you lean on real captures ($0) — like this walkthrough
  • a still with ffmpeg motion (push-in, glow) replaces an i2v shot
  • the leader/outro bracket is built once and reused on every film
  • you validate one PoC before fanning out — no wasted generations
  • the audio is a generated bed and re-runs of the edit stay free
Same bar, different structureThis exact film, made fully-AI (a dozen i2v shots instead of real screens), would land closer to $4–8 — same quality bar, a very different cost. Choosing where to spend generation is the whole game.
Read-out

The studio travels to any brand

Adopt the brand, prove the signature once, drive the real product, fan out for free. A finished, on-brand walkthrough for a second product — in a few hours, at a bar you set, for under a euro of generation.

See it liveelementals.nl — the real product the film is made from. And the studio itself: the video studio guide.