VerifAI
Marketing media

Coded scenes for facts. Diffusion for atmosphere.

The VerifAI plan for explainer video and infographic production: a reference gallery, an honest tooling survey, five Verify-branded storyboards, and one rule — text lives in code, atmosphere lives in pixels.

01

Text in code.

Wordmark, captions, numbers, labels — every legible pixel is coded (SVG/DOM or Remotion), never diffusion-generated.

02

Atmosphere in pixels.

Diffusion video/image only for text-free backgrounds and motion. Composited as a layer, never load-bearing.

03

Brand rules hold.

White background, one green accent for verified, checkmark wordmark locked top-left, no dark, no gold, no gradient blobs.

Storyboards

Five Verify-branded pieces. 45–90 seconds each.

Every storyboard follows the three-step line: A system runs. It produces evidence of what it did. Anyone can check that evidence later.

Storyboard AFlagship, homepage hero loop

How VerifAI works

  1. 01Open on a single clean action — a request, a click — rendered as a coded UI card.
  2. 02The card animates into an AI-system icon (abstract, no internals), processing pulse.
  3. 03A second card: cryptographic receipt icon with timestamp and a 'view proof' affordance.
  4. 04Cut to a simple public verification screen: paste an ID, get a green check.
  5. 05End card: wordmark, tagline 'powered by Veriton,' 'Request a Demo' CTA.
Reference leanServiceNow's calm narrated pacing; Anthropic's short, no-filler demo cuts.
Storyboard BConcierge / site-builder capability

The AI-controlled homepage

  1. 01Split screen: left is a chat panel; someone types a request in natural language.
  2. 02Right is a live site canvas that updates — section reorders, color swap, new card.
  3. 03Small honest label: 'shown as a capability walkthrough' — no live claim.
  4. 04End on the entity-hub pattern as the underlying content model.
Reference leanPalantir AIP screen-recording; ServiceNow's narrated agent workflow.
Storyboard CAI-run company product

Run your company with an AI team

  1. 01Open on a coded org chart. Each node is an agent role, not a person.
  2. 02A task enters at the top, flows through 2–3 nodes, each stamped with a verification tick.
  3. 03Dashboard card: '12 agents, 340 tasks this week, 100% auditable' (illustrative).
  4. 04End card: 'this is how Veriton itself runs' honesty line, CTA.
Reference leanCognite's complexity-to-simplicity arc; Palantir architecture-explainer graphics.
Storyboard DWorks & Demos flagship

Concierge / entity-hub demo

  1. 01Open on the 3-zone entity-hub layout: links left, identity+chat middle, actions right.
  2. 02Cursor clicks a left link; middle updates. Cursor clicks a right action; panel opens.
  3. 03Overlay coded labels once each: 'LINKS' left, 'ACTIONS' right.
  4. 04End card: 'built once, reused everywhere' line, CTA.
Reference leanCognite / Palantir real-product screen-recording honesty. No invented footage.
Storyboard EStretch, if time allows

Verifiable ML, not just LLMs

  1. 01Open on a simple training-loop icon (coded, abstract, no internals).
  2. 02A small tape/ledger icon accumulates ticks as the loop advances.
  3. 03Cut to an auditor screen: 'replay this run' button, byte-identical result badge.
  4. 04End card, CTA.
Reference leanAnthropic 'computer use' pacing — calm, procedural, no hype cuts.
Tooling survey

Six generators. One pipeline we already own.

Universal finding: no current AI-video generator reliably renders legible on-brand text or a logo inside the frame. Lead with the coded pipeline; use diffusion only for text-free B-roll.

Runway Gen-4.5
Access
runwayml.com, web/API
Cost
$12–76/mo, ~12 credits/sec
Text / logo in-frame
Weak — add in post
Pro
60s continuous 4K, built-in editor
Con
Highest per-second cost of the four
Google Veo 3.1
Access
Gemini API / Flow / Google One
Cost
$0.03–0.40/sec API; $19.99–249.99/mo
Text / logo in-frame
Weak — official guidance: do not rely on it for text
Pro
Cheapest API tier, native audio
Con
Explicit official warning against in-frame text/logos
Kling AI 3.0
Access
klingai.com, web/API
Cost
Free (watermarked) to $90–100/mo Ultra
Text / logo in-frame
Best of the four, still imperfect
Pro
Only tool with usable in-frame titles
Con
Paid tier needed to remove watermark
Pika 2.5
Access
pika.art, web/API
Cost
Free to $76/mo; ~$0.05/sec API
Text / logo in-frame
Weak; image-to-video keeps a supplied logo image consistent
Pro
Fast, cheap; strong at holding a supplied logo/product image steady
Con
Direct text-in-prompt mostly fails
Synthesia
Access
synthesia.io
Cost
Free (watermarked) to $29–89/mo; Ent. ~$30k/yr
Text / logo in-frame
Brand Kit applies logo/colors consistently (paid)
Pro
Only tool with a real brand-kit + watermark-removal system
Con
Minute caps are tight — a few 30-second videos/month
Coded HTML → Chromium → FFmpeg pipeline
Access
In-house, already built
Cost
$0 marginal
Text / logo in-frame
Perfect — every pixel of text is coded
Pro
Full control of brand, motion, captions; zero drift risk
Con
More manual scene-building than one-prompt generators
Recommendation

Lead with the coded HTML → Chromium → FFmpeg (or Remotion) pipeline for anything with real product UI, numbers, or the wordmark on screen — zero brand-drift, $0 marginal cost. Use Kling 3.0 or Pika (image-to-video from a still of the check-badge) only for short text-free atmosphere B-roll. Synthesia's Brand Kit is a once-a-month tool for an avatar-narrated cut, not a production pipeline.

Infographic plan

Same rule, still pictures.

GPT-image tests around 99% character-accurate; Nano Banana around 94% and cheaper. Neither publishes a brand-constrained benchmark, so palette-holding is unproven. Render text and structure in code; use AI only for the atmosphere layer.

How it works (3-step: runs, evidence, verify)
Coded SVG diagram, no AI image
Too fact-dense for AI text rendering
Product-catalog capability map (12 × 16)
Coded grid/table, no AI image
A data grid should never be a diffusion image
How the AI-controlled homepage works
AI background (text-free, green tints) + coded SVG on top
Keep the AI layer decorative only
Start your own AI-run company org chart
Coded SVG chart, no AI image
Structured data — a diffusion image would hallucinate labels
AI-models logo wall
Coded logo grid with real vendor marks
Logos must be exact; a licensing/attribution question, not generation
Compliance timeline (EU AI Act, Aug 2 2026, etc.)
Coded SVG timeline, no AI image
Dates and article numbers must be exact
General atmosphere / hero backgrounds
GPT-image or Nano Banana, text-free, brand-constrained
Manual palette review before compositing
See also
How it works →
The public three-step flow this media illustrates.
See also
Works & Demos →
The Sognefjord concierge and other demos referenced in Storyboards B and D.
Talk to us
[email protected] →
Bring a storyboard, we'll ship a first cut.