CLIENT NORTHCOUNT (INVENTED BRAND) · SAAS PRODUCT FILM · 16:9, 3 FEATURE CLIPS, 4:5, 1:1, 9:16, 4 SITE LOOPS, STEMS · 1:03 · 2026
Product film forNorthcount (concept)
A 63-second film for a revenue analytics product: the interface built and animated in 2D, every action on the word that names it, cut for every feed.
Northcount is an invented product and brand, made for this portfolio. Every company, account and figure on screen is fictional, and the interface, charts and logo were built for this video. What was generated is listed at the end.
northcount-film-16x9.mp4 · 1920×1080 · 30 fps · 1:03 · −14.2 LUFS integrated, measured on the delivered file
The brief
We're launching Northcount, revenue analytics for subscription businesses: it explains every change in recurring revenue with the accounts that moved it. We need a product film of about 75 seconds (16:9) that someone who has never heard of us understands: what it does, who it's for and why it matters. Voice-over, kinetic type and the real interface in 2D: dashboards, data, charts that draw, rows that re-rank, a cursor that clicks and types, notifications. Clean and readable. Then three short feature clips without voice, LinkedIn cuts in 4:5, 1:1 and 9:16 with burned-in captions, and four website loops for the feature sections, with no visible jump at the seam. One campaign: every format from the same plan. Every on-screen text has to be readable at the size each format is seen at, including the LinkedIn feed on a phone. Nothing in the picture may come from an image or video model. English only.
Process
Write the idea before building anything.
The script came first, then the central idea on one page, then the product: a fictional company whose accounts, figures and week are computed from one data file, so the waterfall, the 61% and the weekly brief always agree. The brand was built for video: a near-white canvas, jade as the only colour, red only for what was lost, and one typeface with true tabular figures.
THE PIECE IN ONE SENTENCE
A number falls, and the product opens it up to the accounts that moved it.
FIVE ACTIONS ON THEIR WORD
The last point of the week drops on “fell”, and the headline writes itself: Revenue fell $18,400 this week.
A click on the drop opens “Why MRR fell” on the word “drop”.
The waterfall draws on “gained”, and the losses finish growing on “lost”.
The accounts re-rank on “lines up”; two cancellations turn out to be 61% of the losses.
A renewal alert arrives on “watching” and goes to the right person.
Build the product, then record it.
Northcount runs nowhere else. The interface is a working web app with every state reachable by URL and by real clicks, recorded frame by frame on a virtual clock, at twice the size for the shots the camera pushes into. Every recording was captured twice and compared pixel by pixel: all identical.
Cast one voice for the brand.
A man in his late thirties, warm and unhurried, someone in finance explaining to a colleague, cast over three measured rounds. Each section was voiced on its own, and every word timed by forced alignment, snapped to the measured attack of each word. Transcribed back, the four voiced edits match the script with 0.0% word error.
Hang every action on its word.
One plan per deliverable, where every interface action is timed from the word that names it: the action starts 0.1 to 0.3 s before the word and the result lands on it, and the cursor arrives 0.45 s before each click. The dashboard camera moves a few times and then holds so the screen can be read, framed so every edge of the frame falls in white space.
Compose each format for its screen.
Every format has its own composition of the same screens, recorded from a focus view drawn for its aspect ratio, so each cut is its own recording. Text that is read measures at least 36.5 px in capitals at 1080 wide in the 4:5, 1:1 and 9:16 cuts, 19.8 minutes of arc on a phone, with figures at 72 px and captions at 44 px, 48 px in the 9:16. What it does, who it's for and why it matters are said and written by 4.7 s in every cut.
ONE STATE IN FOUR FORMATS · 16:9, 4:5, 1:1, 9:16 · SCROLL
Loops that close on their own first frame.
Four silent loops for the product's website, one per function, in MP4 and WebM at 0.81 to 1.09 MB each. The last frame is the first frame, both encoded as keyframes from identical pixels: 0.0 levels of difference at the seam in all eight files. They ship with a spec and the embed code for whoever builds the site.
Put the sound on the action.
The music was built from its stems so the bass and drums enter on the cut to the dashboard, the voice sits 9.6 LU over it, and each interface action gets one short sound inside the image that justifies it: 21 in the film, with 0.0 s of sound outside its image.
Look at every frame.
Strips of twelve frames at every cut, every text and every state change, and close-ups at three times the size. Two rounds of fixes, each made in the plan and rendered again: camera edges moved into white space, function labels moved off the app's navigation bar, the camera arriving before the rows re-rank, captions split where the voice breathes, and each loop's seam taken to zero after the codec.
CONTACT SHEET · 42 FRAMES · ONE EVERY 1.5 S · SCROLL
Final and variants
| File | What | Size | fps | Length | Audio |
|---|---|---|---|---|---|
| northcount-film-16x9.mp4 | the product film, the full script | 1920×1080 | 30 | 62.87 s | AAC 48 kHz stereo |
| northcount-feature-01-explain-16x9.mp4 | feature clip, Explain any change, without voice | 1920×1080 | 30 | 17.17 s | music and interface sound |
| northcount-feature-02-ask-16x9.mp4 | feature clip, Ask in plain English, without voice | 1920×1080 | 30 | 17.17 s | music and interface sound |
| northcount-feature-03-alerts-16x9.mp4 | feature clip, Renewal alerts, without voice | 1920×1080 | 30 | 17.17 s | music and interface sound |
| northcount-linkedin-4x5.mp4 | the main LinkedIn cut, burned-in captions | 1080×1350 | 30 | 26.87 s | AAC 48 kHz stereo |
| northcount-linkedin-1x1.mp4 | square cut, composed on its own | 1080×1080 | 30 | 26.87 s | AAC 48 kHz stereo |
| northcount-vertical-9x16.mp4 | vertical cut, captions in the safe zone | 1080×1920 | 30 | 20.03 s | AAC 48 kHz stereo |
| northcount-linkedin-4x5-12s.mp4 | short 4:5 cut for awareness | 1080×1350 | 30 | 12.77 s | AAC 48 kHz stereo |
| northcount-loop-*.mp4 and .webm | four silent website loops, last frame equal to the first, with their spec and embed code | 1600×1000 | 30 | 6.03 to 7.83 s | none |
| northcount-film-16x9_*.wav | stems: voice, music, interface sounds, music and effects, and the mix | 48 kHz, 24-bit | 62.88 s each | stereo | |
| northcount-film-16x9.srt and .vtt | captions timed to the spoken words | ||||
| *-cover.jpg | a cover frame for each video, with the hook headline |
The MP4s are H.264 High, yuv420p, BT.709, 30 fps, with the index at the front for instant playback. Every format comes from the same plan, the same voice and the same type layer, and the loops play muted on any page.
Numbers
Measured on the delivered files.
| check | film 16:9 | feature clips | 4:5 and 1:1 | 9:16 |
|---|---|---|---|---|
| loudness (target −14 LUFS ±1) | −14.2 LUFS | −14.0 LUFS | −14.2 LUFS | −14.2 LUFS |
| true peak (max −1 dBTP) | −1.8 dBTP | −1.4 to −1.7 dBTP | −1.8 dBTP | −1.9 dBTP |
| voice over music, while speaking | +9.6 LU | no voice | +9.7 LU | +10.0 LU |
| interface sounds outside their image | 0.0 s, 21 sounds | 0.0 s, 6 to 7 sounds | 0.0 s, 6 sounds | 0.0 s, 6 sounds |
| smallest cap height of text that is read | 25.3 px at 1x; figures 44.7 px, panel titles 55.1 px | 36.5 px | 36.5 px; figures 72.3 px, captions 44 px | 36.5 px; figures 72.3 px, captions 48.4 px |
| worst text contrast (min 4.5) | 5.68:1 | 5.68:1 | 5.68:1 | 5.68:1 |
| what it does, who for, why it matters: said and written by | 6.6 s (limit 8) | n/a | 4.7 s (limit 5) | 4.7 s (limit 5) |
| the voice, transcribed back against the script | 0.0% word error | no voice | 0.0% word error | 0.0% word error |
| frames counted against length × fps | 1886 of 1886 | 515 of 515, each | 806 of 806, each | 601 of 601 |
Action on the word: 51 interface state changes are tied to the word that names them. 44 start on it, within −0.3 to +0.15 s (median −0.10 s; the worst, −0.32 s, is the question box lighting up one frame early), and 7 finish on it within ±0.15 s. 26 are measured on the recorded interface frame by frame; the gradual ones (counts, typing, masks) are timed from the plan.
Outside the references, by choice: interface on screen 84.6% of the film (the product films we measured: 55 to 65%, where the rest is a person on camera; here the voice carries the context), and 3.5 camera moves every 20 s (5 to 6), so each screen holds long enough to read: the camera stays still 53.1% of the time in holds of 1.2 s or more.
One campaign: one typeface and the same five motion curves in every file. The brand's jade and loss red measure within ΔE2000 1.8 of their tokens in six of the eight videos; in the Explain clip they measure 2.10 and 2.75, and in the Ask clip the jade measures 4.65, against a limit of 2.0. In the three long cuts the ink measures 9.6 to 10.2, where the caption strip, ink at 85% over the canvas, enters the measure. didn't pass
The loops: 0.0 levels of difference between the last and the first frame in all eight files, 0.81 to 1.09 MB each, with no audio track. The stems line up with the mix to the sample, offset 0. Every interface recording was captured twice and compared pixel by pixel: all identical.
A comprehension test is written and proposed: 10 to 15 people who have never seen the product watch the first 8 seconds of the film with sound, or the first 5 of the 4:5 cut muted, and write down what it does, who it is for and why they would want it. It has yet to be run.
No views, clicks or conversions: the film has not run.
What we did
Script, the product and its brand, the interface, direction, edit, motion, sound and the casting of the voice. The interface recordings, the type layer, all twelve deliverables, the mix and the loops were made in FireNgine, our production system.
Provenance
Generated: the voice. The rest is licensed or made in FireNgine.
Source files
Available on request: the edit of each deliverable as a timeline your own editor can open, and the loops' spec with its embed code.
Have a script, a voice-overor a product to launch?
Send a short brief or write to us. We'll come back with questions, then one central idea for the piece before anything is made.
hello@firendev.com