Silver Fern · Creative Library · Program Scope · One Page
Teaser 5 ("how a rush order happens," 34s, shipped 7/4) proves the format works: fernie UI theater, kinetic type, a booth end card, a music bed. This page is the whole video program on one sheet — what ships now, what ships cheap, and what we wait on — so nobody has to scope it two minutes at a time.
Proves: the library's motion language survives the jump to real MP4. The rush-order story lands in 34 seconds with zero live footage — UI theater and type beats carry it. Missing, in order of impact: a voiceover (the audio is music-only; the A1 voice-clone from the sales pipeline already exists and is cleared — narration is a solved problem we haven't plugged in), and the fernie face on the end card is an off-model draw, not the canonical symbol — the character bible (Vol. 9) now exists precisely so every appearance uses the same face.
| Tier | What | Cost / when |
|---|---|---|
| 1 · RENDER FARM | Every motion card in the library becomes an MP4. ~500 LOOP/SETTLES cards across Vols 2–14, screen-recorded headless at native resolution, vidpipe audio bed, −14 LUFS, exported 16:9 + 9:16 + 1:1. The library's per-card PNGs, but for video — hundreds of ready social/booth clips with zero new creative decisions. Fully mechanical; the pipeline pieces (headless Chromium, ffmpeg, vidpipe) all exist. | Near-zero marginal cost. Ships this week on a say-so. |
| 2 · THE 30-SECOND MACHINE | Teaser 5, industrialized. A repeatable film format: hook line → UI theater beat → payoff type → booth/product end card, 30–45s, voiceover from the cleared voice-clone, canonical fernie only. Each film is a one-page brief (story, beats, claims) run through the same gates as the library (brand law, character law, fact pack). First slate: one film per module (5), the consolidation story (1), the fernie origin (1), plus re-cutting teaser 5 with VO and the canonical face. | Cheap per film once the template exists. First slate in days, not weeks. |
| 3 · AI FOOTAGE | The "getting closer" part — generated live-action-style shots for what CSS can't do: greenhouse b-roll, the 4AM dock, hands and soil, glass at dawn. The Cathedral volume (Vol. 7) is already the storyboard — 48 epic frames waiting to move. Standing rules before the first frame ships: AI footage never depicts the product UI (real UI gets comped in from Tier 2), never depicts real or realistic people's faces, every clip through a media-safety gate like /hmvs, human sign-off per film. | Don't build yet. Run a monthly one-hour sample: generate 3 Cathedral shots on the current best model, keep the results in a dated folder. Ship the first real film the month the samples stop looking generated. |
Kevin's instinct — "not sure if it's worth the time investment, but it's getting closer" — splits cleanly: Tiers 1–2 are worth it now because they spend almost nothing and multiply assets that already passed the gates. Tier 3 is a calendar entry, not a project — the monthly sample run costs an hour and tells us exactly when the threshold crosses, without betting real time on a moving target. Nothing here needs a new team or a new tool purchase.