Use a small, silent video to make a shared link feel alive. Start with one test page, verify playback, then invest in an original 3D loop.
Recommended first step
A half-day delivery experiment for Godfrey: a simple original 2D loop, a still poster and a separate share URL. If current Messages clients play it reliably, produce a 4–6 second Godfrey 3D loop and a distinct 2–3 second OI loop.
This report proposes options. Neither product site was changed and no new film was produced.
What the example actually does
The reference post shows a Messages-style link card for Check Supply. Paper checks move through a small cream 3D machine with green-rimmed transparent tubes. The camera stays fixed. The title and domain sit below the picture, so the animation needs no tiny text.
2 seconds
The actual preview asset, not the 6.485-second recording of the card.
253 KB
Silent H.264 MP4, 1200 × 630, 30 fps. The separate poster is about 101 KB.
Its initial HTML declares a direct og:video MP4 plus an og:image fallback. It still uses twitter:card=summary_large_image. That demonstrates a compact rich-link setup, not a universal video card across social apps.
Measured from the live page and public video asset. Source frames were inspected privately and are not reproduced here. Borrow the clear repeating action; make original artwork.
Where the two sites stand
Raw HTML and public media checked with browser, Slackbot and Twitterbot user agents. This tests delivery, not app playback.
Observed
Godfrey
Operational Intelligence
Current share preview
Static image and summary_large_image. No video, player or oEmbed tags found.
Static image and summary_large_image. No video, player or oEmbed tags found.
Website hero
Image in the initial HTML.
Existing muted autoplay video, 17.33 seconds and 13.41 MB.
Media issue to address
Keep the existing poster as a fallback while testing a separate video.
Export a much smaller social clip. Current hero serves video/x-m4v; a preview should serve video/mp4. Move the poster off the legacy site hostname when updating it.
Implementation confidence
Local source confirmed: Vinext/Vite/React with Next-style metadata and market-specific IN/US variants. Inspect emitted tags rather than assume full Next.js parity.
Live metadata and media verified. The authoritative homepage source repository was not confirmed; identify it before assigning an exact code change.
Public surfaces: Godfrey and OI. A hero video on a webpage and a social link preview are independent placements.
Three production options
All three use the same delivery contract: an MP4, a useful poster and server-rendered metadata. Fallbacks are required whichever creative direction we choose.
Process motion
Large paper forms, a clear work queue and one exception moving to review. Crisp 2D shapes and typography explain the service at small sizes.
Planning estimate: 2–4 person-days for both brands, including a revision and exports. Fastest to edit; easiest to make generic. Best for testing the transport before commissioning detailed models.
Physical 3D loop Creative recommendation
One believable mechanism fills the card. Paper thickness, material edges and contact shadows carry the realism. Use a fixed or nearly fixed camera, continuous motion and no miniature office full of unreadable details.
Planning estimate: 3–5 person-days for a lean pair of loops using available scene foundations. Allow 5–8 if geometry, lighting and revision work become custom. Strongest resemblance to the reference’s craft, with original objects and meaning.
3D plus crisp process labels
One tactile rendered scene, supported by a short, readable 2D label. This can be clearer for a service unfamiliar to the viewer. Put the important context in the link title too, so the static card still makes sense.
Planning estimate: 3–5 person-days for both brands if scenes stay simple. More text introduces localization and crop work. Reserve it for concepts that fail the silent thumbnail test without explanation.
These are rough human production estimates, not quotes or guarantees. They include one revision round. Custom people, complex simulation, licensed models and additional review rounds increase effort. The small delivery prototype is separate: roughly half a working day, with device access available.
How the reviewers reached a shared recommendation
The 3D reviewer favored a Godfrey tabletop review cycle and an OI shared operating mechanism. The independent video reviewer preferred process motion as the first investment because playback support was still unknown.
The synthesis is staged: use a minimal 2D clip to prove delivery, then select original physical 3D as the finished visual direction. A poster remains useful even where video does not play.
3D review was assigned to GPT-6 Astra. Video/editorial and platform reviews used Codex. Claude Opus 5.5 did not run; an available-model review was used instead.
Two brands, two recognizable ideas
Godfrey
Work reaches review
Use a close, fixed view of three shallow trays. Cream paper, muted green felt, restrained walnut edges. One amber-tabbed exception stays available for an accountant. The message is organized financial work with professional judgment.
Concept sketch only. The proposed film would use physical 3D objects, not this diagram.
First frame
A tidy work surface already shows the input, prepared folder and separate review item. It must work as a poster.
One 4–6 second cycle
A sheet arrives, aligns with a record, then enters the draft tray. The exception remains visible.
Loop and title
A new sheet enters as the last leaves. Suggested card title: “Financial expertise, backed by AI”. No speed counter or autonomous-payment claim.
Operational Intelligence
Distinct businesses, shared operating support
Three substantial service modules share one recessed track and visible mechanical drive. Small work tiles enter, emerge aligned and continue. Charcoal ceramic, brushed metal and ivory distinguish this from Godfrey’s paper workflow.
Concept sketch only. The metaphor needs brand approval before modeling; it does not assert an actual shared software product.
First frame
Three distinct modules and their common support are visible immediately. Do not turn OI into an accounting-only brand.
One 2–3 second cycle
A continuous work tile moves through the mechanism. Keep the camera fixed at a restrained three-quarter angle.
Loop and title
The mechanism returns seamlessly. Suggested card title: “Building the next generation of services”. Avoid invented dashboards or customer results.
Brand interpretation grounded in OI’s public thesis. Exact positioning and final copy should be approved with the selected concept.
The delivery plan
Original sceneOne action with a useful first frame
Render onceSilent H.264 MP4 and still poster
Share pageRaw HTML metadata and stable public URLs
Real client testRecord motion, image fallback and failures
No live 3D runtime is needed for the card. Crawlers fetch the rendered files. A separate website hero can reuse approved media later, but has its own loading, playback and accessibility decisions.
Documented capabilities and the remaining uncertainty. No new product clip has been tested inside these apps.
Consumer
Primary evidence
Plan for this project
Messages
Apple’s rich-link guidance describes direct-file og:video, inline looping playback and initially muted audio. It gives a 10 MB combined icon/image/video budget.
Best first target. This is older official guidance, not a freshly measured universal cutoff. Verify current iOS/macOS, connection and privacy settings.
Plan for an image. No direct-file video guarantee was verified; Meta documentation could not be fully retrieved in this audit.
Proposed asset contract and metadata
Purpose-built silent loop, 1200 × 630, H.264 MP4, yuv420p, fast-start metadata. Aim for roughly 0.3–2 MB and a poster under 300 KB. These are design targets, not platform limits.
Public HTTPS assets, correct video/mp4 / image MIME types, successful unauthenticated GET and HEAD, and byte-range GET returning 206. No login, consent wall, expiring signatures or bot challenge.
Versioned media filenames with long-lived immutable caching. Keep HTML metadata updateable and account for app-side caches.
Render OG tags in the initial head. Preserve title, description, canonical URL, locale and the image fallback. Godfrey’s IN/US variants must not accidentally inherit one another’s share copy.
Illustrative direct-file metadata, not code applied to either site:
<meta property="og:type" content="website">
<meta property="og:title" content="Financial expertise, backed by AI">
<meta property="og:url" content="https://example.com/share/review">
<meta property="og:image" content="https://example.com/social/review-v1.jpg">
<meta property="og:image:alt" content="Prepared records with an item set aside for review">
<meta property="og:video" content="https://example.com/social/review-v1.mp4">
<meta property="og:video:secure_url" content="https://example.com/social/review-v1.mp4">
<meta property="og:video:type" content="video/mp4">
<meta property="og:video:width" content="1200">
<meta property="og:video:height" content="630">
<meta name="twitter:card" content="summary_large_image">
See the Open Graph specification. In Godfrey’s inspected source, the likely entry point is a page-specific Next-style openGraph.videos value; validate actual Vinext output. OI needs its authoritative source and deployment path confirmed before implementation.
Cost, mobile, accessibility and privacy
Cost: production and revisions dominate this small asset. No paid generation or asset purchase is required for the proposed pilot. At 1 MB per file, 1,000 full uncached downloads would transfer about 1 GB; crawler caches and range fetches change actual usage. Confirm the existing hosting plan before assigning a recurring cost.
Mobile and accessibility: the poster and link title must carry the idea without motion, tiny type or audio. A later on-site video should have controlled loading, a pause control for persistent motion, and a reduced-motion poster. External apps control their own preview playback and preferences.
Privacy and security: use original geometry and synthetic records without customer names, balances, account numbers or internal screens. A future iframe or oEmbed endpoint would need its own narrow security design; neither is required for this first experiment.
A small first milestone
Prove delivery. Make one simple original 2D loop and poster for a dedicated Godfrey share URL. Keep the existing homepage unchanged. This is the proposed next implementation step, not something built for this report.
Check the bytes and HTML. Confirm early metadata, public media, correct MIME, range responses, a clean full decode, small combined payload and a readable poster at about 320 pixels wide.
Measure actual client behavior. Share the fresh URL on current Messages devices. Record OS/app, screenshot or recording, autoplay, looping and fallback. Test additional social apps only within the authorized target destinations.
Choose the creative finish. If playback is dependable, build the short Godfrey 3D review loop and then the OI operating mechanism. If an important client remains static, keep the useful poster and avoid claiming video support.
Roll out deliberately. Only after a concept and placement are selected, use the normal reviewed GitHub-main deployment process. Replacing either homepage hero is a separate decision.
Acceptance means observed playback and a good fallback
The clip explains one action silently, its seam is unobtrusive, the first frame works as an image, and the test log names the clients that played it. A validator or a fetched tag alone is not a playback result.
Sources and confidence
Michael Pfister’s reference post: text recovered through X’s public oEmbed and syndication endpoints; decoded recording frames inspected. Live reference assets probed separately.
Check Supply: direct metadata and asset measurements, checked 24 October 2026.
Measured: current HTML, public asset headers, file sizes, codecs and dimensions. Documented: platform capabilities above. Proposed: concepts, effort and engineering targets. Unverified: playback of any future Godfrey/OI clip inside real sharing clients and the live OI homepage source repository. No private source messages or customer records are included.